# Dreamdrive

> Dreamdrive is a read-only Chrome extension that saves every image and video a person makes with AI, together with the prompt, model and settings that made it, in one searchable library sorted by client. It is made by Synthetic White AB in Stockholm, Sweden. Dreamdrive is not in the Chrome Web Store yet; there is a waiting list at https://www.dreamdrive.ai/ and one email goes out when it is ready to install.

Site: https://www.dreamdrive.ai/ · Contact: info@dreamdrive.ai · Company: Synthetic White AB, Stockholm, Sweden (org. no. 559519-7665)
This file: https://www.dreamdrive.ai/llms.txt · Full text: https://www.dreamdrive.ai/llms-full.txt · HTML version: https://www.dreamdrive.ai/ai/

## What Dreamdrive is

### What is Dreamdrive?
Dreamdrive is a read-only Chrome extension that saves every image and video a person makes with AI, together with the prompt, model and settings that made it, in one searchable library sorted by client. It is made by Synthetic White AB in Stockholm, Sweden. Dreamdrive is not in the Chrome Web Store yet; there is a waiting list at https://www.dreamdrive.ai/ and one email goes out when it is ready to install.

### Does Dreamdrive generate images or run anything on my behalf?
No. Dreamdrive only reads: it never clicks, types, submits, queues or generates anything on any platform. It records what you make, with the settings you used, and nothing else.

### Who is Dreamdrive for?
Freelancers, studios, agencies and in-house teams who make images and video with AI for clients and need to say afterwards which prompt, model and tool made a given result. It is also used by people who simply want to find the picture they made three months ago.

## Platforms

### Which AI platforms does Dreamdrive work with?
Dreamdrive works with 38 named platforms: Midjourney, ChatGPT, Gemini, Adobe Firefly, Leonardo, Runway, Higgsfield, Sora, Google Flow / Whisk, Kling, Luma Dream Machine, Pika, Hailuo, Vidu, Wan, LTX Studio, Hedra, HeyGen, invideo AI, Ideogram, Flux (BFL Playground), Krea, Recraft, Freepik AI, Magnific, Dzine, Dreamina, Canva AI, Grok, Stability AI, Civitai, Tensor.Art, SeaArt, NightCafe, Replicate, fal, ComfyUI, Stable Diffusion web UI. Files from any other tool can be imported, and the prompt is read from the file when it is embedded.

### Which platforms are captured automatically?
Every web platform you connect (36 today) is captured automatically once you switch it on: Midjourney, ChatGPT, Higgsfield, Kling, Runway and the rest. Midjourney is the most precise. Local tools (ComfyUI and Stable Diffusion web UI) work through file import, which reads the prompt and settings embedded in the files.

### Does Dreamdrive capture video as well as images?
Yes. On video platforms (Runway, Higgsfield, Sora, Google Flow / Whisk, Kling, Luma Dream Machine and more) Dreamdrive keeps the prompt, the model, the duration and resolution, and the clip itself. Videos play inside the Dreamdrive viewer.

## How capture works

### How does Dreamdrive capture a generation?
You switch a platform on in the Connect list; Chrome asks whether Dreamdrive may read that site. From then on each generation you make there appears in Dreamdrive's side panel a moment later with its prompt, model, settings, references and results. Nothing is read on sites you have not switched on.

### What if a site changes and something is missed?
Hover or right-click any image on any site and choose Save to Dreamdrive, or import the files. Fields saved by hand are marked manual, fields read from a file are marked file, and fields read from the page are marked auto.

### What are the four ways to get something into Dreamdrive?
Automatic capture on a connected platform, hover-save, right-click Save to Dreamdrive on any image, and file import. All four land in the same library.

## What is stored

### What does Dreamdrive store for each capture?
The platform, the prompt, the model, the settings, the references, every result (image or video), the time, the project and folder you chose, the approval status and any notes. Each field is marked auto, file or manual to show how it was obtained.

### Are files kept at full resolution?
On the Pro plan every image and video is stored at the resolution the platform delivered it. The Free plan keeps the last 90 days of captures with thumbnails; older captures are kept, not deleted, and become visible again on Pro.

### Where is my data stored?
Captures are stored on servers in the European Union by Synthetic White AB, Stockholm. Prompts, client names and notes are used to run your library and for nothing else: not for training, not for marketing.

## Pricing

### How much does Dreamdrive cost?
The Free plan costs $0 for ever and keeps the last 90 days of your work with 3 projects. Pro costs $9 a month or $79 a year (about $6.60 a month) and adds the full history, unlimited client projects, full-resolution files, the one-page project summary, share links and downloads. Prices are in US dollars and every account starts with 14 days of Pro free, with no card.

### What happens when a trial or subscription ends?
The account becomes Free. Nothing is deleted: the projects and the full history are there again the moment you return to Pro.

## Privacy

### Is Dreamdrive private?
Yes. Captures are private to your account and nothing is shared unless you create a share link, which is unguessable, not indexed by search engines and revocable. Dreamdrive only reads the signed-in user's own work on the platforms they switched on.

### Is Dreamdrive affiliated with Midjourney, OpenAI, Google or any other platform?
No. Dreamdrive is an independent product of Synthetic White AB. Platform names and logos appear only to say where Dreamdrive works and belong to their owners.

## Hand-over and comparisons

### What is the project summary?
A one-page summary per client project: per platform the models used, your declared plan and a link to the platform's terms, then every approved capture with its prompt, settings and references. It prints, exports to PDF or CSV, and can be shared through a revocable link. It lists facts and is not legal advice.

### How is Dreamdrive different from a notes app or a screenshot folder?
Dreamdrive records automatically while you work, keeps the prompt and settings attached to each result, marks how every field was obtained, and sorts everything by client. A notes app or a folder depends on you remembering to save and to write things down.

### How is Dreamdrive different from the platforms' own history?
Each platform keeps only its own history, in its own place, and loses the client context. Dreamdrive keeps every platform in one library, with projects, approvals and a summary you can hand over.

## Platform pages

- [Dreamdrive for Midjourney](https://www.dreamdrive.ai/for/midjourney/)
- [Dreamdrive for ChatGPT](https://www.dreamdrive.ai/for/chatgpt/)
- [Dreamdrive for Gemini](https://www.dreamdrive.ai/for/gemini/)
- [Dreamdrive for Adobe Firefly](https://www.dreamdrive.ai/for/firefly/)
- [Dreamdrive for Leonardo](https://www.dreamdrive.ai/for/leonardo/)
- [Dreamdrive for Runway](https://www.dreamdrive.ai/for/runway/)
- [Dreamdrive for Higgsfield](https://www.dreamdrive.ai/for/higgsfield/)
- [Dreamdrive for Sora](https://www.dreamdrive.ai/for/sora/)
- [Dreamdrive for Google Flow / Whisk](https://www.dreamdrive.ai/for/google-labs/)
- [Dreamdrive for Kling](https://www.dreamdrive.ai/for/kling/)
- [Dreamdrive for Luma Dream Machine](https://www.dreamdrive.ai/for/luma/)
- [Dreamdrive for Pika](https://www.dreamdrive.ai/for/pika/)
- [Dreamdrive for Hailuo](https://www.dreamdrive.ai/for/hailuo/)
- [Dreamdrive for Vidu](https://www.dreamdrive.ai/for/vidu/)
- [Dreamdrive for Wan](https://www.dreamdrive.ai/for/wan/)
- [Dreamdrive for LTX Studio](https://www.dreamdrive.ai/for/ltx/)
- [Dreamdrive for Hedra](https://www.dreamdrive.ai/for/hedra/)
- [Dreamdrive for HeyGen](https://www.dreamdrive.ai/for/heygen/)
- [Dreamdrive for invideo AI](https://www.dreamdrive.ai/for/invideo/)
- [Dreamdrive for Ideogram](https://www.dreamdrive.ai/for/ideogram/)
- [Dreamdrive for Flux (BFL Playground)](https://www.dreamdrive.ai/for/flux/)
- [Dreamdrive for Krea](https://www.dreamdrive.ai/for/krea/)
- [Dreamdrive for Recraft](https://www.dreamdrive.ai/for/recraft/)
- [Dreamdrive for Freepik AI](https://www.dreamdrive.ai/for/freepik/)
- [Dreamdrive for Magnific](https://www.dreamdrive.ai/for/magnific/)
- [Dreamdrive for Dzine](https://www.dreamdrive.ai/for/dzine/)
- [Dreamdrive for Dreamina](https://www.dreamdrive.ai/for/dreamina/)
- [Dreamdrive for Canva AI](https://www.dreamdrive.ai/for/canva/)
- [Dreamdrive for Grok](https://www.dreamdrive.ai/for/grok/)
- [Dreamdrive for Stability AI](https://www.dreamdrive.ai/for/stability/)
- [Dreamdrive for Civitai](https://www.dreamdrive.ai/for/civitai/)
- [Dreamdrive for Tensor.Art](https://www.dreamdrive.ai/for/tensorart/)
- [Dreamdrive for SeaArt](https://www.dreamdrive.ai/for/seaart/)
- [Dreamdrive for NightCafe](https://www.dreamdrive.ai/for/nightcafe/)
- [Dreamdrive for Replicate](https://www.dreamdrive.ai/for/replicate/)
- [Dreamdrive for fal](https://www.dreamdrive.ai/for/fal/)
- [Dreamdrive for ComfyUI](https://www.dreamdrive.ai/for/comfyui/)
- [Dreamdrive for Stable Diffusion web UI](https://www.dreamdrive.ai/for/a1111/)

## Pages

- [Home](https://www.dreamdrive.ai/)
- [Get started: install to hand-over in seven steps](https://www.dreamdrive.ai/get-started/)
- [All platforms](https://www.dreamdrive.ai/for/)
- [Guides](https://www.dreamdrive.ai/guides/)
- [Midjourney glossary](https://www.dreamdrive.ai/glossary/)
- [Legal](https://www.dreamdrive.ai/legal/)
- [Privacy policy](https://www.dreamdrive.ai/privacy-policy/)
- [Terms](https://www.dreamdrive.ai/terms/)

## Guides

- [How to write a Midjourney prompt (structure + examples)](https://www.dreamdrive.ai/guides/midjourney-prompt-structure/): The anatomy of a strong Midjourney prompt: subject, composition, style, lighting and parameters like --ar, --stylize and --sref, with copy-ready examples.
- [Midjourney --sref: how to get a consistent style](https://www.dreamdrive.ai/guides/midjourney-style-references-sref/): What --sref style references and codes are, how to tune them with --sw, and how to keep a reusable library of looks for a whole project.
- [How to organise your Midjourney prompts](https://www.dreamdrive.ai/guides/organize-midjourney-prompts/): A simple system for saving, naming and reusing Midjourney prompts by project, so the prompt that worked last week is one click away.
- [Midjourney Parameters Explained: Full Reference Table](https://www.dreamdrive.ai/guides/midjourney-parameters-explained/): Every common Midjourney parameter in one reference table: what --ar, --s, --c, --w, --seed, --no, --sref and more do, and exactly when to use each.
- [Midjourney Prompt Generator: Build Better Prompts Faster](https://www.dreamdrive.ai/guides/midjourney-prompt-generator/): What a good Midjourney prompt generator should do, an honest comparison of hand-writing, chatbots, template lists and Dreamtune, and how Dreamtune works.
- [Midjourney Aspect Ratios: The Right --ar for Every Platform](https://www.dreamdrive.ai/guides/midjourney-aspect-ratios/): Which Midjourney --ar to use for Instagram feed and Stories, YouTube, LinkedIn and X banners, Pinterest and print, plus how aspect ratio changes composition.
- [Midjourney Camera Angles: Prompt Words and What They Convey](https://www.dreamdrive.ai/guides/midjourney-camera-angles/): Every Midjourney camera angle from eye level to worm's-eye and dutch angle: what each one communicates, and the exact prompt wording that gets it.
- [Midjourney Lighting Prompts: Terms, Effects and Examples](https://www.dreamdrive.ai/guides/midjourney-lighting-prompts/): Midjourney lighting prompts explained: golden hour, softbox, Rembrandt, rim, high-key, neon and chiaroscuro, with the effect of each and copy-ready examples.
- [Midjourney Product Photography Prompts: 13 Copy-Ready Examples](https://www.dreamdrive.ai/guides/midjourney-product-photography-prompts/): 13 copy-ready Midjourney product photography prompts for packshots, hero and lifestyle images, plus surfaces, lighting and the limits on label text.
- [Midjourney portrait prompts: formula + 12 examples](https://www.dreamdrive.ai/guides/midjourney-portrait-prompts/): A reliable Midjourney portrait prompt formula with lighting, framing, skin texture and respectful depiction tips, plus 12 copy-ready portrait prompts.
- [How to Make Consistent Characters in Midjourney](https://www.dreamdrive.ai/guides/midjourney-consistent-characters/): Practical techniques for keeping the same character across Midjourney images: fixed descriptions, seeds, image prompts, style references and character sheets.

---

# Midjourney glossary

**Prompt**: The text description Midjourney turns into images. A strong prompt names the subject, composition, style, lighting and environment, followed by any parameters.

**Parameter**: An option added to the end of a Midjourney prompt, written with two dashes, such as --ar 16:9. Parameters change how an image is generated without changing the description.

**Aspect ratio (--ar)**: The width-to-height ratio of the image, for example --ar 16:9 for widescreen or --ar 4:5 for an Instagram portrait post.

**Stylize (--stylize, --s)**: Controls how strongly Midjourney applies its own default aesthetic. Low values follow your words more literally; high values are more artistic.

**Chaos (--chaos, --c)**: Controls how different the images in a single grid are from one another. Higher chaos gives more varied, less predictable results.

**Weird (--weird, --w)**: Pushes results toward unusual, unconventional aesthetics. Useful for experimental looks.

**Seed (--seed)**: The number Midjourney uses as the starting point for an image. Reusing a seed with the same prompt and settings gives more repeatable results.

**No (--no)**: A negative prompt: tells Midjourney what to leave out, for example --no text, watermark.

**Style raw (--style raw)**: Reduces Midjourney's default beautification so results follow the prompt more literally. Popular for photography and precise art direction.

**Style reference (--sref)**: Transfers the visual style of a reference image or style code (colour, texture, lighting, mood) to a new image without copying its content.

**Style code**: A number used with --sref, such as --sref 2980181076, that stands for a particular look. Style codes are easy to save, share and reuse for consistent series.

**Style weight (--sw)**: Sets how strongly a --sref style reference influences the image. Lower is subtle; higher pushes the image closer to the reference style.

**Image prompt**: An image URL placed at the start of a prompt. Unlike --sref, an image prompt influences both the content and the style of the result.

**Image weight (--iw)**: Sets how much an image prompt matters relative to the text prompt.

**Version (--v)**: Chooses which Midjourney model version generates the image. Results, parameter ranges and style codes can differ between versions.

**Niji (--niji)**: A Midjourney model tuned for anime and illustrative styles.

**Tile (--tile)**: Generates images that repeat seamlessly, for patterns, textures and wallpapers.

**Personalization (--p)**: Applies your personal aesthetic, which Midjourney learns from images you rate or like, to a prompt.

**Upscale**: Increases an image's resolution and detail after generation.

**Vary (Subtle / Strong)**: Creates new variations of an existing image, either close to the original (Subtle) or more different (Strong).

**Dreamtune**: Dreamdrive's prompt builder. It turns a rough idea plus dropdown choices (composition, subject, reference, technical, style, environment) into a complete Midjourney prompt.

**My Prompts**: Dreamdrive's prompt library. It saves every prompt automatically into project folders that you can search and reuse with one click.

**Style Vault**: Dreamdrive's style library. It stores your --sref style codes with preview images and includes 1,000+ curated styles.

**Credit (Dreamdrive)**: The unit Dreamtune uses: 1 credit generates 1 prompt. New accounts get 5 free credits; subscriptions include a monthly allowance and Credit Packs top up.

---

# How to write a Midjourney prompt (structure + examples)

URL: https://www.dreamdrive.ai/guides/midjourney-prompt-structure/
Updated: 2026-10-02

Short answer: A good Midjourney prompt describes, in order, the subject, the composition (shot type, angle, framing), the style and medium, the lighting and environment, and ends with parameters such as --ar for aspect ratio or --sref for a style reference. Be specific about what you want to see, and keep each idea short.

The five building blocks

Subject. Who or what is in the image, with the details that matter: a man in a navy business suit holding a coffee cup.

Composition. Shot type, camera angle, framing and depth: medium shot, low angle, centred, shallow depth of field.

Style and medium. Photograph, oil painting, 3D render, film still, illustration, plus any era or movement.

Lighting and environment. Where it happens and how it is lit: rainy Tokyo street at night, neon reflections, soft rim light.

Parameters. Settings that go at the very end, each starting with two dashes.

Parameters worth knowing

Parameter | What it does | Example |

--ar | Aspect ratio of the image | --ar 16:9 |

--stylize / --s | How strongly Midjourney applies its own aesthetic (higher = more artistic, lower = closer to your words) | --s 250 |

--chaos / --c | How different the images in one grid are from each other | --c 20 |

--no | Things to leave out | --no text, watermark |

--sref | Use a style reference image or style code | --sref 2980181076 |

--seed | Reuse a starting point to get more repeatable results | --seed 1234 |

--style raw | Less default beautification, more literal results | --style raw |

Parameter ranges and defaults can change between model versions, so check Midjourney's own documentation when something behaves unexpectedly.

Examples

Weak: businessman, cool, cinematic

Strong: a man in a navy business suit crossing a rainy Tokyo street at night, medium shot, low angle, shallow depth of field, cinematic film still, neon reflections on wet asphalt, soft rim light --ar 16:9 --s 200

The strong version tells Midjourney what to show, how to frame it, what it should look like and how it is lit. Every word is doing a job.

Common mistakes

Stacking adjectives like beautiful, stunning, amazing. They add little; concrete details add a lot.

Contradictions such as minimalist, highly detailed, busy.

Burying parameters in the middle of the prompt. Put them at the end.

Not saving what worked. The prompt that nailed it last week is the one you can't find today.

Doing this faster with Dreamdrive

Dreamdrive's Dreamtune turns these building blocks into dropdowns (Composition, Subject, Reference, Technical, Style and Environment), so you pick camera angle, framing, shot type, focus and depth instead of remembering the vocabulary. Every result is saved to My Prompts automatically.

Q: Where do parameters go in a Midjourney prompt?
A: At the end of the prompt, after the text description. Each one starts with two dashes, for example --ar 16:9 --s 200.

Q: How long should a Midjourney prompt be?
A: Long enough to describe subject, composition, style and lighting clearly, usually one or two sentences. Specific short phrases beat long lists of adjectives.

Q: What does --stylize do in Midjourney?
A: It controls how strongly Midjourney applies its own aesthetic. Lower values follow your words more literally, higher values are more artistic.

---

# Midjourney --sref: how to get a consistent style

URL: https://www.dreamdrive.ai/guides/midjourney-style-references-sref/
Updated: 2026-10-02

Short answer: --sref tells Midjourney to copy the visual style (colour, texture, lighting, mood) of a reference image or a numeric style code, without copying its content. Add --sref followed by an image URL or a code such as --sref 2980181076, and use --sw to control how strongly that style is applied.

Image URL or style code

There are two ways to give Midjourney a style reference:

An image URL. Midjourney reads the look of that image and applies it to your prompt.

A style code. A number like 2980181076 that stands for a particular look. Codes are easy to share and reuse, which makes them ideal for series work and brand consistency.

Typing --sref random gives you a random style, and the job shows the code that was used, so you can keep the ones you like.

Controlling strength with --sw

--sw (style weight) sets how much the reference influences the result. Lower values give a light touch; higher values push the image strongly toward the reference. Start at the default and move in steps until the balance between your prompt and the style feels right.

Combining styles

You can list more than one reference after --sref to blend looks, and add weights with :: to favour one over another, for example --sref 1111111111::2 2222222222::1.

Keeping looks consistent across a project

Collect. Each time you get a style you like, save its code with a preview image.

Name and group. Organise styles by client, project or mood so you can find them later.

Note the model version. Style codes can render differently on different Midjourney versions, so record which version a code was made on.

Reuse the same code and weight for every image in a series.

Doing this faster with Dreamdrive

Dreamdrive's Style Vault is built for exactly this: save --sref codes next to their preview images, sort them into collections, favourite the ones you reuse, and browse a library of 1,000+ curated styles when you need a starting point.

Q: What is the difference between --sref and an image prompt?
A: An image prompt influences both content and style. --sref transfers only the style: colour, texture, lighting and mood, not the objects in the image.

Q: Where do I find Midjourney style codes?
A: Use --sref random to discover new ones, note codes from your own jobs, or browse curated libraries such as the Dreamdrive Style Vault.

Q: What does --sw mean in Midjourney?
A: --sw is style weight. It controls how strongly the --sref reference is applied to your image.

---

# How to organise your Midjourney prompts

URL: https://www.dreamdrive.ai/guides/organize-midjourney-prompts/
Updated: 2026-10-02

Short answer: Save every prompt at the moment you make it, group prompts by project rather than by date, keep the parameters and style codes with the prompt, and make your library searchable. A browser extension like Dreamdrive does this automatically inside midjourney.com.

Why prompts get lost

Midjourney keeps a history of your jobs, but over hundreds of generations it becomes a long feed sorted by time. When a client comes back for "the same look as before", you end up scrolling, or rebuilding the prompt from memory and never quite matching it.

A system that works

Save at creation, not later. If saving is a separate step, it won't happen on a busy day.

Organise by project. One folder per client, campaign or series. Dates are a poor index for creative work.

Keep the full prompt. Include every parameter and style code. The --ar, --s and --sref values are often what made it work.

Name variations clearly. Man in a business suit, ... green suit, ... red suit reads better than v2_final_final.

Make it searchable. A library you can't search is a library you won't use.

Keep styles separate. Store reusable looks (style codes) in their own collection so they can be mixed with any subject.

Options compared

Method | Good for | Drawback |

Midjourney job history | Finding recent work | Sorted by time, not by project |

Notes app or spreadsheet | Free and flexible | Manual copy-paste; easy to forget |

Dreamdrive My Prompts | Automatic saving into project folders, inside midjourney.com | Requires a Dreamdrive subscription |

Doing this with Dreamdrive

My Prompts saves every prompt automatically, lets you create projects, search, rename and delete, and sends any saved prompt back to Midjourney with one click. Pair it with the Style Vault for your reusable looks.

Q: Does Midjourney save my prompts?
A: Midjourney keeps your job history, which includes prompts, but it is ordered by time rather than organised by project.

Q: What is the best way to store Midjourney prompts?
A: Save them automatically as you create them, grouped by project, with all parameters and style codes kept intact, in a searchable library.

---

# Midjourney Parameters Explained: Full Reference Table

URL: https://www.dreamdrive.ai/guides/midjourney-parameters-explained/
Updated: 2026-10-02

Short answer: Midjourney parameters are settings you add to the end of a prompt with a double hyphen, such as --ar 16:9 or --s 250. They control things the words can't, like aspect ratio, how strongly Midjourney applies its own aesthetic, how varied the grid is, and which model or style reference is used.

How Midjourney parameters work

A parameter is an instruction to the image generator rather than a description of the image. Your words describe what you want to see; parameters change how Midjourney produces it.

The rules are simple:

Parameters go at the end of the prompt, after all your descriptive text.

Each starts with two hyphens, followed by the name, a space and usually a value: --ar 3:2.

Separate parameters with spaces. The order between them doesn't matter.

Don't put commas or full stops after a parameter. Punctuation can be read as part of the value and cause an error.

Many parameters have a short form. --stylize 250 and --s 250 do the same thing.

A complete prompt with parameters looks like this:

a lighthouse on a rocky coast at dusk, long exposure, moody blue light --ar 3:2 --s 150 --c 10

If you're still working on the descriptive part, our guide to Midjourney prompt structure covers how to order subject, style and setting before you add any parameters.

The Midjourney parameter reference table

This table covers the parameters most people use day to day. Ranges and defaults can change between model versions, so check Midjourney's own docs if a value is rejected.

Parameter | Short form | What it does | Typical values | Use it when |

--aspect | --ar | Sets the width-to-height ratio of the image | Whole numbers, e.g. 1:1, 4:5, 16:9, 9:16 | You know where the image will be used |

--stylize | --s | Controls how strongly Midjourney applies its default aesthetic | 0-1000, default 100 | Images look too plain or too "Midjourney" |

--chaos | --c | Increases variety between the images in a grid | 0-100, default 0 | You're exploring ideas and want different takes |

--weird | --w | Pushes results towards unconventional, offbeat looks | 0-3000, default 0 | You want something surprising or experimental |

--seed | none | Fixes the starting noise so results are repeatable | A whole number | Testing one change at a time |

--no | none | Negative prompt: tells Midjourney what to leave out | One or more words, comma-separated | An unwanted element keeps appearing |

--style raw | none | Reduces Midjourney's automatic beautification | raw | You want literal, photographic or precise results |

--sref | none | Applies a style reference from an image URL or numeric code | URL, code, or random | You want a consistent look across images |

--sw | none | Sets how strongly the style reference applies | 0-1000, default 100 | A style reference is too weak or too dominant |

--iw | none | Sets how much an image prompt influences the result | Check docs for your version | You're using an image prompt at the start |

--tile | none | Creates a seamless repeating pattern | No value needed | Textures, wallpapers, fabric prints |

--niji | none | Switches to the anime and illustration model | A version number | Anime, manga or stylised illustration |

--version | --v | Chooses the model version | A version number | You need an older model's look or features |

--personalize | --p | Applies your personalisation profile | No value, or a profile code | You want results nudged towards your taste |

--repeat | --r | Runs the same prompt several times | A small whole number | You want lots of options from one prompt |

Parameters that shape composition and variety

--ar (aspect ratio)

The aspect ratio changes the canvas shape, and Midjourney composes for that shape. A portrait at --ar 9:16 is framed differently from the same prompt at --ar 16:9, so set it from the start rather than cropping later. Ratios must use whole numbers. Our aspect ratio guide lists the right ratio for each platform.

--chaos and --weird

--chaos makes the four images in a grid differ more from each other. --weird makes each image stranger. They solve different problems: chaos is for exploring, weird is for character. Low values of both are often enough.

--seed

Every job starts from random noise. A seed number fixes that noise, so the same prompt with the same seed and settings gives very similar results. It's the best tool for testing one change at a time.

Parameters that control style

--stylize

Low values follow your prompt more literally. High values give Midjourney more freedom to add its own sense of composition, colour and detail. The default of 100 is a sensible middle.

--style raw

Raw mode tones down the built-in polish. It's useful for photography prompts, product shots and anything where you want the words to matter more than Midjourney's taste.

--sref and --sw

A style reference copies the look of an image (colour, texture, mood) without copying its content. You can use an image URL or a numeric style code, and --sref random picks a random code for you to discover new looks. --sw controls the strength. Style codes may render differently across model versions, so test old codes after an update. See our style references guide for the details.

--niji, --v and --p

--niji uses a model tuned for anime and illustration. --v picks a model version. --p applies your personalisation profile, built from images you've ranked or liked, so results lean towards your taste.

Parameters for image prompts and special outputs

Image prompts: paste an image URL at the start of the prompt, then add --iw at the end to control its influence. Higher values follow the image more closely.

--no: removes unwanted elements, e.g. --no text, watermark. Write the things you don't want as plain nouns.

--tile: makes the edges match so the image repeats seamlessly. Try hand-drawn lemons and leaves, flat pastel illustration --tile.

A practical example combining several:

ceramic coffee mug on a linen tablecloth, soft window light, minimal product photography --ar 4:5 --style raw --s 50 --no text, logo

Here --style raw and a low --s keep the product realistic, and --no stops Midjourney inventing a printed logo. Remember that Midjourney is poor at rendering long, exact text, so add real typography in a design tool afterwards.

Common parameter mistakes

Putting parameters in the middle. Anything after a parameter can be misread. Keep all of them at the end.

Using a single hyphen or an en dash. It must be two plain hyphens. Some apps autocorrect -- into a dash, so check when pasting.

Decimal aspect ratios. --ar 1.91:1 won't work. Use whole numbers, e.g. --ar 191:100.

Changing several parameters at once. You won't know which one made the difference. Change one, keep the seed fixed, compare.

Writing "no" in the prompt text. "A beach, no people" can add people, because "people" is in the prompt. Use --no people instead.

Maxing everything out. --s 1000 --c 100 --w 3000 rarely gives usable images. Start low and raise values gradually.

Doing this faster with Dreamdrive

Dreamtune builds prompts from dropdowns, including a Technical group, so you can pick settings without typing them from memory. My Prompts saves every prompt, parameters included, into searchable project folders for one-click reuse. Style Vault keeps your --sref codes with preview images.

Q: Where do parameters go in a Midjourney prompt?
A: At the very end, after all your descriptive text. Each starts with two hyphens, for example --ar 16:9 --s 200.

Q: Does the order of Midjourney parameters matter?
A: No. As long as all parameters sit at the end of the prompt, you can list them in any order.

Q: What is the default stylize value in Midjourney?
A: The default is 100 on a range of 0 to 1000. Lower values follow your prompt more literally; higher values add more of Midjourney's own aesthetic.

Q: Why is my Midjourney parameter not working?
A: Common causes are a single hyphen or autocorrected dash instead of two hyphens, punctuation after the value, decimal aspect ratios, or a parameter not supported by the model version you're using.

Q: What's the difference between --chaos and --weird?
A: --chaos makes the images in a grid more different from each other. --weird makes each image more unconventional and strange.

---

# Midjourney Prompt Generator: Build Better Prompts Faster

URL: https://www.dreamdrive.ai/guides/midjourney-prompt-generator/
Updated: 2026-10-02

Short answer: A good Midjourney prompt generator gives you a clear structure, precise vocabulary to choose from, correct parameters at the end of the prompt and a way to save what works, while leaving the output as editable text. Chat assistants are good for ideas, template lists for learning, and a builder inside Midjourney, such as Dreamdrive's Dreamtune, for consistent, repeatable prompts.

What a good Midjourney prompt generator should do

A prompt generator is any tool or method that helps you turn an idea into a well-formed Midjourney prompt. The good ones are not about producing long, impressive-sounding text. They make the right decisions visible so you stop forgetting them. Judge any generator against these five jobs.

Give you structure. A strong prompt covers subject, setting, composition, lighting, style and finish in a sensible order, with parameters at the end. A generator should walk you through those slots so nothing important is missing. The prompt structure guide explains the order.

Put vocabulary at your fingertips. Most weak prompts are weak because the writer did not know the word: "rim light", "three-quarter view", "overcast diffused daylight", "tilt-shift". A generator should offer precise terms you can choose from rather than expecting you to remember them.

Handle parameters correctly. It should place --ar, --s, --chaos, --no, --style raw and --sref at the end, with valid syntax, and never invent parameters that Midjourney does not have. See parameters explained for what each one does.

Save what you make. The prompt behind a great image is an asset. If the generator does not save it, or you do not, you will struggle to recreate the result.

Keep you in control. The output should be plain, editable text that you can read and adjust, not a black box.

It also helps if the tool sits close to where you actually generate, because copying between tabs is where prompts get lost and parameters get mangled.

What a generator cannot do for you

No generator guarantees a good image. Midjourney still interprets every prompt, and the same text can produce a strong grid and a weak one. A generator cannot fix Midjourney's known weaknesses either: long exact text, logos and precise product details remain unreliable whatever the prompt says.

The best generators also do not replace taste. You still decide which result fits the brief, which image to vary and when to stop. Think of a generator as a fast, well-organised assistant for the writing part, not a substitute for looking.

Be wary of very long generated prompts. Piling on twenty style words often muddies the result, because terms compete with each other. A focused prompt usually beats a long one:

a lighthouse on a rocky coast, stormy sea, dramatic clouds, golden hour rim light, wide shot, documentary landscape photography --ar 16:9 --style raw

Comparing your options honestly

There are four common ways to build prompts. Each has a place, and many people mix them.

Approach | Strengths | Weaknesses | Best for |

Writing by hand | Full control, free, teaches you how Midjourney responds | Slow, easy to forget parameters, inconsistent wording between sessions | Experienced users and one-off experiments |

Chat assistants (general AI chatbots) | Flexible, good for brainstorming ideas and rewording, works from a loose description | Can suggest outdated or non-existent parameters, tends towards wordy prose, means copying between tabs, no built-in prompt library | Early ideas and concept exploration |

Template and prompt lists | Free, quick to copy, a good way to learn vocabulary | Generic results that many people share, need editing to fit your subject, can go out of date | Learning and quick inspiration |

A builder inside Midjourney, such as Dreamtune | Structured slots and curated vocabulary in one place, works where you generate, prompts can be saved and reused | Choices are limited to the options offered, needs desktop Chrome, paid after the free credits | Regular users who want consistent, repeatable prompts |

A sensible mix: brainstorm the idea however you like, use a structured builder to turn it into a clean prompt, then edit by hand once you see the first grid.

How Dreamtune works

Dreamtune is the prompt builder in Dreamdrive, a Chrome extension that works inside midjourney.com in desktop Chrome. You can read the full overview in what Dreamdrive is.

Dropdowns in six groups. Options are grouped into Composition, Reference, Technical, Styles, Subjects and Environments, so you build a prompt slot by slot rather than from a blank box.

References. The Reference group covers artists, photographers, eras, movies, fashion brands and games, which saves you hunting for the right term.

Simple or Complex output. Choose a short, focused prompt or a richer, more detailed one.

Credits. You get 5 free credits on sign-up, and 1 credit is one Dreamtune prompt. See Dreamdrive pricing for plans.

The result is ordinary prompt text, so you can still read it, tweak a word or add a parameter before generating. Two related features sit alongside it: My Prompts saves every prompt into searchable project folders, and Style Vault stores your --sref codes with preview images and gives you a library of over 1,000 curated styles. Both need a subscription.

Getting more out of any generator

Whichever approach you use, these habits make the output better:

Start with a clear subject. Generators fill in style; they cannot guess what the picture is about.

Change one thing at a time. Swap only the lighting, or only the lens, so you learn which choice made the difference.

Add a style reference for consistency. Pairing a generated prompt with the same style reference keeps a set coherent.

Run the grid more than once. Judge a prompt on two or three runs, not one lucky or unlucky grid.

Keep the winners. Save the exact prompt and parameters next to the image you approved.

Here is the kind of prompt a structured builder helps you reach, with each slot filled deliberately:

a ceramicist shaping a vase at a wheel, close-up on clay-covered hands, warm window light from the side, shallow depth of field, quiet documentary photography, muted earth tones --ar 4:5 --style raw --s 150

Doing this faster with Dreamdrive

Dreamtune builds prompts from dropdowns in Composition, Reference, Technical, Styles, Subjects and Environments, with Simple or Complex output, right inside midjourney.com. My Prompts keeps every prompt in searchable project folders for one-click reuse, and Style Vault holds your --sref codes. New accounts get 5 free credits to try Dreamtune.

Q: What is the best Midjourney prompt generator?
A: It depends on how you work. Chat assistants are flexible for brainstorming, template lists are free for learning, and a builder inside Midjourney such as Dreamtune gives structured, repeatable prompts without switching tabs.

Q: Can ChatGPT or other chatbots write Midjourney prompts?
A: Yes, they are useful for ideas and rewording, but they can suggest outdated or non-existent parameters and tend to write long prose, so check the parameters before you run the prompt.

Q: Is Dreamtune free?
A: You get 5 free credits on sign-up, and each credit makes one Dreamtune prompt. My Prompts and Style Vault need a subscription.

Q: Does a prompt generator guarantee better images?
A: No. It helps you write complete, well-structured prompts, but Midjourney still interprets each prompt differently and remains unreliable at exact text and product details.

---

# Midjourney Aspect Ratios: The Right --ar for Every Platform

URL: https://www.dreamdrive.ai/guides/midjourney-aspect-ratios/
Updated: 2026-10-02

Short answer: Set the aspect ratio in Midjourney with --ar followed by width and height in whole numbers, for example --ar 4:5 for an Instagram feed post or --ar 9:16 for Stories and Reels. The default is square (1:1), and Midjourney composes the image for the ratio you choose, so set it before you generate rather than cropping afterwards.

How --ar works

The aspect ratio is the relationship between an image's width and height. In Midjourney you set it with --aspect or its short form --ar at the end of the prompt:

a cyclist crossing a misty bridge at dawn, cinematic wide shot --ar 16:9

Three things to know:

Width comes first. --ar 16:9 is landscape; --ar 9:16 is portrait.

Whole numbers only. --ar 1.91:1 won't work, but --ar 191:100 will.

The ratio changes the composition, not just the crop. Midjourney arranges the subject, background and empty space to suit the canvas. A tall ratio gives room for a full-length figure; a wide one invites landscapes and side-by-side layouts.

Without --ar, you get a square image.

Aspect ratios by platform and use

Platforms change their recommended sizes from time to time, so treat this as a starting point and check the platform's own guidance for pixel dimensions.

Use | Ratio | Midjourney parameter | Notes |

Instagram feed, portrait | 4:5 | --ar 4:5 | Takes up the most screen space in the feed |

Instagram feed, square | 1:1 | --ar 1:1 | Midjourney's default; safe for grids and profile layouts |

Instagram Stories and Reels | 9:16 | --ar 9:16 | Keep key details away from the top and bottom, where buttons sit |

TikTok and YouTube Shorts | 9:16 | --ar 9:16 | Same full-screen vertical format |

YouTube thumbnail and video frame | 16:9 | --ar 16:9 | Leave space if you'll add a title in a design tool |

YouTube channel banner | 16:9 | --ar 16:9 | Only a central strip shows on every device; keep the subject centred |

LinkedIn profile banner | 4:1 | --ar 4:1 | Very wide; the profile photo covers part of the left side |

X (Twitter) header | 3:1 | --ar 3:1 | Keep the lower left clear for the profile photo |

Pinterest pin | 2:3 | --ar 2:3 | Tall pins stand out in the grid |

Website hero or blog header | 16:9 or 21:9 | --ar 21:9 | Wide formats leave room for overlaid headlines |

Desktop wallpaper | 16:9 | --ar 16:9 | Use 21:9 for ultrawide monitors |

Phone wallpaper | 9:16 or taller | --ar 9:19 | Many modern phones are taller than 9:16 |

Classic photo print | 3:2 or 2:3 | --ar 3:2 | Matches 4x6 inch prints |

8x10 inch print | 4:5 | --ar 4:5 | Same shape as an Instagram portrait post |

5x7 inch print or card | 5:7 | --ar 5:7 | Greeting cards and invitations |

A-series paper (A4, A3, A2) | about 1:1.414 | --ar 70:99 | A close whole-number match for ISO paper sizes |

Cinematic still | about 2.39:1 | --ar 239:100 | Widescreen film look; pairs well with "film still" wording |

Designing for the format

Getting the ratio right is only half the job. The prompt should suit the shape too.

Vertical formats (9:16, 4:5, 2:3)

Describe things that are naturally tall: a full-length figure, a tall building, a waterfall, a bottle shot from slightly below. Mention where the subject sits, e.g. "subject in the lower third, open sky above" if you need room for text.

full-length portrait of a woman in a long camel coat walking down a narrow cobbled street, overcast light --ar 9:16

Wide formats (16:9, 3:1, 4:1, 21:9)

Describe horizontal scenes and use space deliberately. For banners, ask for the subject on one side and "negative space" on the other.

a single sailing boat on the far right of a calm grey sea, vast empty sky, minimal, plenty of negative space on the left --ar 4:1

Square (1:1)

Square suits centred, symmetrical subjects: a product, a face, a logo-like icon. Note that Midjourney isn't reliable at rendering long, exact text, so design the wording of any logo or banner separately.

Choosing between similar ratios

Some ratios are close enough that the choice comes down to how the image will be used.

4:5 or 1:1 for Instagram? 4:5 takes up more vertical space in the feed, which suits people, products and fashion. 1:1 is safer if the image also needs to work as a square thumbnail elsewhere.

2:3 or 4:5 for portraits? 2:3 gives more room above and below the subject, useful for full-length figures and posters. 4:5 is tighter and suits head-and-shoulders framing.

16:9 or 21:9 for headers? 16:9 is the safer all-rounder. 21:9 feels more cinematic but leaves less height for the subject, so it works best for landscapes and wide scenes.

3:2 or 16:9 for landscape photos? 3:2 feels like a classic camera frame. 16:9 feels like a video still.

When in doubt, generate at the ratio of the most important placement, then create separate versions for the others using the same prompt and seed so the family of images still feels consistent.

Practical tips for aspect ratios

Decide the destination first. If an image needs to work as both a 4:5 post and a 9:16 story, generate each separately rather than cropping one.

Use Pan or Zoom Out to extend. If you love an image but need a different shape, Midjourney's pan and zoom-out tools can extend the canvas in a direction.

Be careful with extreme ratios. Very wide or very tall ratios can produce stretched or repeated elements. If results look odd, try a slightly less extreme ratio and crop.

Upscale before printing. Midjourney's output size depends on the ratio and model. For large prints, upscale first and check the pixel dimensions against your printer's requirements.

Keep the ratio when testing. Changing --ar changes the composition, so keep it fixed when comparing other settings like stylize, chaos and weird.

For the full list of settings that sit alongside --ar, see Midjourney parameters explained.

Doing this faster with Dreamdrive

Dreamtune includes Composition and Technical dropdowns, so you can choose framing and format settings without remembering the syntax. My Prompts saves each prompt with its ratio into project folders, which makes it easy to rerun a campaign in every format you need.

Q: What aspect ratio should I use in Midjourney for Instagram?
A: Use --ar 4:5 for portrait feed posts, --ar 1:1 for square posts and --ar 9:16 for Stories and Reels.

Q: What is Midjourney's default aspect ratio?
A: The default is square, 1:1. Add --ar with a width and height, such as --ar 16:9, to change it.

Q: Can I use decimals in Midjourney --ar?
A: No. Aspect ratios must use whole numbers. For 1.91:1, write --ar 191:100.

Q: What aspect ratio is A4 in Midjourney?
A: A-series paper is about 1:1.414. --ar 70:99 is a close whole-number match for A4 and other ISO sizes.

Q: Does changing the aspect ratio change the image?
A: Yes. Midjourney composes the scene for the canvas shape, so the same prompt and seed can look quite different at 9:16 and 16:9.

---

# Midjourney Camera Angles: Prompt Words and What They Convey

URL: https://www.dreamdrive.ai/guides/midjourney-camera-angles/
Updated: 2026-10-02

Short answer: To set a camera angle in Midjourney, name it with the standard photography term near the start of your prompt, such as low-angle shot, looking up at, and describe what that viewpoint would reveal. Low angles convey power, high angles vulnerability, eye level neutrality and dutch angles unease.

Why camera angle changes the story

Camera angle is where the lens sits in relation to your subject: how high or low it is, and whether it is level or tilted. The same person, building or product can look heroic, fragile, neutral or unsettling purely because of where the camera is placed.

Midjourney has no virtual camera to position. It learned angles from captioned photographs and film stills, so it responds to the phrases photographers and cinematographers actually use. That means two things for you: use the standard names (low-angle shot, not camera looking up a bit), and describe what the angle would make visible, so the model has more than one clue pointing in the same direction.

Angle is separate from shot type. Shot type is how much of the subject fills the frame (close-up, full shot, wide shot). Angle is the viewpoint. You will usually want one of each in a prompt.

Camera angles at a glance

Use this table to pick an angle by the feeling you want, then copy the wording into your prompt.

Angle | What it communicates | Prompt wording to try |

Eye level | Neutral, honest, relatable. The viewer meets the subject as an equal. | eye-level shot, camera at eye level |

Low angle | Power, confidence, heroism, scale. The subject looms over the viewer. | low-angle shot, looking up at |

High angle | Vulnerability, smallness, overview. The viewer looks down on the subject. | high-angle shot, looking down at |

Bird's-eye view | Order, pattern, detachment. Turns scenes into shapes and maps. | bird's-eye view, top-down overhead shot |

Worm's-eye view | Awe and extreme scale. Buildings tower, people become monuments. | worm's-eye view from ground level |

Dutch angle | Unease, tension, disorientation. Something is wrong. | dutch angle, tilted horizon |

Over-the-shoulder | Conversation, connection, a shared point of view. | over-the-shoulder shot |

Point of view (POV) | Immersion. The viewer becomes the character. | first-person POV shot, own hands visible |

Aerial | Geography and scale of a whole landscape or city. | aerial drone photograph |

Three-quarter view | Dimensional and flattering. Shows the face and its shape. | three-quarter view |

Profile | Contemplative, formal, graphic silhouette. | side profile view |

Rear view | Mystery, anonymity, a journey ahead. | shot from behind, back to camera |

The last three describe which way the subject faces rather than the camera's height, but Midjourney treats them as viewpoint cues and they combine well with the height-based angles above.

How to word angles so Midjourney follows them

Angles are easy to lose in a long prompt. These habits make them stick:

Put the angle near the start. Words early in a prompt tend to carry more weight. Low-angle shot of a lighthouse keeper beats burying "low angle" after ten style words.

Pair the name with its direction. low-angle shot, looking up at or high-angle shot, looking down on gives two agreeing signals instead of one.

Describe what the angle reveals. A low angle shows sky and ceilings behind the subject. A high angle shows the floor around them. Adding sky filling the background or wet cobblestones around her feet reinforces the viewpoint.

Match the aspect ratio to the angle. Tall frames such as --ar 9:16 suit worm's-eye views of towers. Wide frames such as --ar 16:9 suit over-the-shoulder and aerial shots.

Try --style raw for photographic work. It reduces Midjourney's default aesthetic, which can make literal camera directions easier to see.

Keep a good composition and explore it. When one image in the grid nails the angle, use Vary (Subtle) to explore details without losing the viewpoint.

A working example that combines these:

low-angle shot looking up at a lighthouse keeper in a yellow oilskin coat, storm clouds filling the sky behind him, rain on the lens, documentary photograph --ar 2:3 --style raw

Choosing an angle by emotional goal

Power, confidence and heroism

Go low. A low angle makes the subject dominate the frame and the viewer feel smaller. Push it further with a worm's-eye view for monuments, architecture and fantasy giants. Wide-angle lens cues exaggerate the effect (see the lens and depth of field guide).

worm's-eye view from ground level of a brutalist concrete tower, converging vertical lines, overcast sky, architectural photography --ar 9:16 --style raw

Vulnerability, intimacy and smallness

Go high. Looking down on a subject makes them seem small or exposed. A high angle on a child, an animal or a lone figure in a big space reads as tender or lonely depending on the lighting.

high-angle shot looking down at a small girl sitting alone on a large empty tiled floor, soft window light, quiet melancholy mood --ar 4:5

Tension and unease

Tilt the camera. A dutch angle is the classic thriller and horror device. If the tilt comes out too subtle, ask for it explicitly with strongly tilted horizon or canted frame.

dutch angle, strongly tilted horizon, a man in a trench coat in a narrow neon-lit alley at night, wet pavement, cinematic film still --ar 21:9

Clarity, design and pattern

Go overhead. A bird's-eye or top-down view flattens depth and turns a scene into a layout. It is the standard for flat lays, food photography, board games and maps. Add photograph if you are getting isometric illustrations instead.

top-down overhead photograph of a breakfast table, eggs, coffee and croissants arranged on a linen cloth, soft morning light --ar 1:1

Connection and conversation

Use an over-the-shoulder shot. The foreground shoulder anchors the viewer in the scene and points their attention at the person or thing being faced.

over-the-shoulder shot of a detective facing a suspect across an interrogation table, single hanging lamp, film noir --ar 16:9

Common problems and how to fix them

The angle is ignored. Move it to the first few words, add the matching direction ("looking up at") and describe what the background should show. Reroll a couple of times before rewriting, because the grid varies.

Faces look stretched at extreme angles. That is how wide lenses close to a face behave in real photos. Use a gentler angle, add a longer lens cue such as 85mm, or step the camera back with a wider shot type.

Overhead shots turn into illustrations. Bird's-eye wording is common in isometric art. Add photograph and a light source, and consider --style raw.

Two angles fight each other. low angle, bird's-eye view is contradictory. Pick one viewpoint per image.

Every result looks the same. Raise --chaos a little to see more varied compositions in the grid, then lock in what works.

Once you have an angle that works, save the exact wording. Small phrasing changes can shift results, and a reliable phrase is worth reusing across projects. For where angle sits in a full prompt, see how to structure a Midjourney prompt.

Doing this faster with Dreamdrive

Dreamtune builds prompts from dropdowns grouped into Composition, Reference, Technical, Styles, Subjects and Environments, so you can assemble a viewpoint, subject and style without retyping, with Simple or Complex output. My Prompts auto-saves every prompt into searchable project folders, so an angle phrase that worked is one click from reuse.

Q: How do I get a low-angle shot in Midjourney?
A: Start the prompt with "low-angle shot, looking up at" followed by your subject, and describe the sky or ceiling behind it. A tall aspect ratio such as --ar 2:3 or 9:16 helps the subject loom.

Q: What is the difference between bird's-eye view and high angle?
A: A high angle looks down on the subject at a slant, so you still see their front. A bird's-eye view looks straight down from directly above, which flattens the scene into a layout.

Q: Why does Midjourney ignore my camera angle?
A: The angle is usually too far back in the prompt or outweighed by other details. Move it to the first few words, add the direction (looking up or down) and describe what the background should show.

Q: What does a dutch angle mean in a prompt?
A: A dutch angle is a tilted camera, so the horizon runs diagonally across the frame. It creates tension or unease, and you can strengthen it with "strongly tilted horizon".

Q: Is camera angle the same as shot type?
A: No. Angle is the viewpoint, such as low or high. Shot type is how much of the subject is in frame, such as close-up or full shot. Most prompts benefit from one of each.

---

# Midjourney Lighting Prompts: Terms, Effects and Examples

URL: https://www.dreamdrive.ai/guides/midjourney-lighting-prompts/
Updated: 2026-10-02

Short answer: To control lighting in Midjourney, name a light source plus its direction, quality and colour, for example soft warm window light from the left or Rembrandt lighting, dark background. Pick one main light and keep the terms consistent, because conflicting lighting words produce muddy results.

Why lighting is your strongest mood control

Change the lighting and you change the whole image. The same face under soft window light feels calm; under a single hard light from the side it feels dramatic; under pink and blue neon it feels like a night out in a city. Lighting words are among the most reliable levers in a Midjourney prompt because they appear constantly in the photo captions and film descriptions the model learned from.

A good lighting phrase answers up to five questions:

Source: sun, window, softbox, candle, neon sign, streetlight.

Direction: from the side, from behind, from above, from below.

Quality: soft and diffused, or hard with crisp shadows.

Colour: warm, cool, golden, teal, magenta.

Contrast: bright and airy (high-key) or dark and shadowy (low-key).

You rarely need all five. Two or three that agree, such as soft warm window light from the left, beat a pile of competing terms.

Lighting terms and their effect

Term | Type | Effect in the image | Use it for |

Golden hour | Natural | Low, warm, directional sun with long shadows and a glow on skin | Portraits, lifestyle, romantic landscapes |

Blue hour | Natural | Cool, even twilight; city lights start to glow | Cityscapes, calm or melancholy scenes |

Overcast light | Natural | Soft, shadowless, muted colours | Honest portraits, fashion, Nordic moods |

Harsh midday sun | Natural | Hard overhead light, deep shadows, high contrast | Summer, desert, graphic fashion |

Window light | Natural | Soft directional light that falls off into shadow | Interiors, still life, intimate portraits |

Backlit | Natural or studio | Light behind the subject, glowing edges, possible lens flare | Dreamy, hopeful, silhouettes |

Softbox lighting | Studio | Large, soft, flattering light with gentle shadows | Product shots, clean portraits |

Rim light | Studio | A bright outline separating the subject from a dark background | Drama, sport, product edges |

Butterfly lighting | Studio | Light high and in front, a small shadow under the nose | Beauty and glamour portraits |

Rembrandt lighting | Studio | Light from one side and above, a small triangle of light on the shadowed cheek | Classic, painterly, moody portraits |

Split lighting | Studio | Half the face lit, half in shadow | Mystery, duality, villains |

High-key lighting | Studio | Bright, low contrast, mostly white tones | Beauty, medical, clean commercial work |

Low-key lighting | Studio | Mostly dark, a few bright areas, strong contrast | Drama, luxury, noir |

Neon lighting | Mood | Saturated coloured light and reflections, often pink, blue or purple | Cyberpunk, nightlife, music |

Chiaroscuro | Mood | Strong contrast between deep darkness and bright light | Painterly drama, Baroque feel |

Candlelight | Mood | Warm, flickering, small pools of light | Historical, intimate, cosy |

Volumetric light | Mood | Visible beams through mist, dust or smoke | Forests, churches, sci-fi interiors |

Coloured gels | Mood | Two or more coloured lights, often contrasting | Editorial, album art, fashion |

Natural light prompts

Natural light terms work best when you also name the time, the weather or the place light comes in. Midjourney has seen countless golden hour photos, so that phrase alone pulls hard towards warm tones; if you want a subtler version, ask for late afternoon light instead.

Golden hour and blue hour

Golden hour is the low sun shortly after sunrise or before sunset. Blue hour is the cool twilight just before sunrise or after sunset. They pair well with backlighting and long shadows.

golden hour backlight, a cyclist on a country lane, warm haze and long shadows, lens flare --ar 3:2 --style raw

blue hour over a harbour town, street lamps just switched on, calm water reflections --ar 16:9

Overcast and window light

Overcast light is like a giant softbox: even, gentle and kind to skin. Window light adds direction, with light on one side of the face and soft shadow on the other.

portrait of a ceramicist in her studio, soft window light from the left, overcast day, muted earthy colours --ar 4:5 --style raw

Studio lighting prompts

Studio terms come from portrait and product photography, so they work best with studio context words such as studio portrait, seamless backdrop or product photograph. Named portrait setups are useful shorthand, though Midjourney treats them as a look rather than an exact diagram, so add the direction too.

Rembrandt: studio portrait of a bearded man, Rembrandt lighting, single key light from the upper left, dark background --ar 4:5

Butterfly: beauty portrait, butterfly lighting from above, glowing skin, soft shadow under the nose, pale grey backdrop --ar 4:5 --style raw

Rim light: sprinter in the blocks, strong rim light outlining the body, black background, sweat catching the light --ar 16:9

High-key: high-key product photograph of a white sneaker on a white seamless background, soft even light, minimal shadows --ar 1:1

Low-key: low-key photograph of a whisky bottle, single hard light from the side, deep black background, amber glow --ar 4:5

For more on product setups, see product photography prompts, and for faces, portrait prompts.

Mood and cinematic lighting

Mood lighting is where colour and contrast do the storytelling. Chiaroscuro and candlelight lean painterly and historical. Neon and coloured gels lean modern, urban and editorial. Volumetric light adds atmosphere and depth, especially with haze, fog or dust in the prompt to give the beams something to show up in.

chiaroscuro oil painting style, an old scholar reading by a single candle, deep darkness around him, warm light on the pages --ar 4:5

neon-lit ramen bar at night, pink and blue signs reflecting on wet pavement, steam rising, cinematic film still --ar 21:9

volumetric light beams through tall pine trees, morning mist, a deer in the clearing --ar 16:9 --style raw

If the result is too dark, add a secondary light such as soft fill light or faint ambient light. If it is too flat, remove words like soft and even and name one strong directional source.

Common lighting mistakes

Stacking conflicting lights. golden hour, neon, high-key, low-key gives Midjourney nothing coherent to follow. Choose one main source and at most one supporting light.

Using "cinematic lighting" alone. It is vague. Say what makes it cinematic: single practical lamp, deep shadows, teal and orange tones.

Forgetting the background. Rim light and low-key looks need a dark background to work. Say so.

Fighting the default aesthetic. Midjourney adds its own polish. For more literal photographic lighting, try --style raw or a lower --stylize value.

Losing a lighting look between images. Once you find one you like, save the exact phrase, or capture the whole look as a style reference.

Doing this faster with Dreamdrive

Dreamtune builds prompts from dropdowns grouped into Composition, Reference, Technical, Styles, Subjects and Environments, with Simple or Complex output. Style Vault stores --sref codes with preview images, so a lighting look you like can be saved and reused alongside a library of 1,000+ curated styles.

Q: What are the best lighting words for Midjourney?
A: Reliable terms include golden hour, blue hour, overcast light, window light, softbox, rim light, Rembrandt lighting, high-key, low-key, neon, chiaroscuro and volumetric light. Add a direction such as "from the left" for more control.

Q: What is Rembrandt lighting in a prompt?
A: It is a portrait setup with one light from the side and slightly above, leaving a small triangle of light on the shadowed cheek. Midjourney reads it as a moody, painterly portrait look.

Q: How do I make Midjourney images darker and more dramatic?
A: Use low-key lighting, a single hard light source from the side and a dark background. Removing words like soft, bright and even also helps.

Q: What is the difference between high-key and low-key lighting?
A: High-key is bright with low contrast and mostly light tones. Low-key is mostly dark with strong contrast and only a few lit areas.

Q: Why does "cinematic lighting" not change much?
A: It is too vague to point to a specific look. Describe the actual light instead, such as a single practical lamp, deep shadows and a colour palette.

---

# Midjourney Product Photography Prompts: 13 Copy-Ready Examples

URL: https://www.dreamdrive.ai/guides/midjourney-product-photography-prompts/
Updated: 2026-10-02

Short answer: A strong Midjourney product photography prompt names the product's material and colour, the surface, the background, the lighting, the camera angle and the purpose of the shot, then ends with parameters such as --ar 4:5 --style raw. Use Midjourney for the scene and composite your real product and label in afterwards, because it cannot reproduce exact products or label text.

The anatomy of a product photography prompt

Product shots live or die on control. A vague prompt such as "a nice photo of a perfume bottle" gives Midjourney too much freedom, and you get a different surface, light and mood on every grid. A reliable product prompt names the same six things every time, in roughly this order, with parameters at the end.

Element | What to specify | Example wording |

Product | Material, colour, shape, finish | a frosted glass serum bottle with a gold dropper cap |

Surface | What it stands on | on a slab of honed travertine |

Background | Colour, texture, depth | plain sage green paper backdrop |

Lighting | Quality and direction | large softbox from the left, gentle fill |

Camera | Angle, lens, focus | eye-level, 100mm macro, sharp focus |

Purpose | The type of image | e-commerce packshot, advertising hero, lifestyle |

Two parameters do a lot of work here. --style raw reduces Midjourney's default artistic interpretation, which usually makes product shots look more like real photography (see style raw). --ar sets the frame, so pick it from where the image will be used rather than leaving everything square.

E-commerce and packshot prompts

Marketplace and shop images need a clean, consistent look: plain backgrounds, even light, the whole product visible and no distracting props. Keep stylisation low and say "catalogue" or "packshot" so the model leans towards plain commercial photography.

a matte black ceramic coffee mug centred on a plain pure white background, soft even studio lighting, subtle contact shadow, e-commerce packshot, sharp focus, high detail --ar 1:1 --style raw

a frosted glass serum bottle with a gold dropper cap on a white infinity curve background, softbox lighting from the left, clean controlled reflections, minimal catalogue photography --ar 4:5 --style raw --no text, props

a pair of white leather trainers at a three-quarter angle on a light grey background, soft shadow beneath, studio product photography, crisp edges --ar 1:1 --style raw

flat lay of a stainless steel wristwatch, spare leather strap and open gift box on a pale beige background, overhead shot, soft diffused daylight, neat even spacing --ar 4:5 --style raw

If stray objects keep appearing, list them after --no. Keep that list short and concrete; see negative prompts for how it behaves.

Hero and advertising prompts

Hero images are where you can be dramatic. You want one strong idea: a striking surface, directional light, some atmosphere and plenty of negative space for a headline added later in design software. Wider frames suit website banners.

a luxury perfume bottle on a polished black marble plinth, dramatic rim lighting, deep shadows, faint mist in the background, premium advertising photography, 100mm macro lens --ar 16:9 --s 250

a cold aluminium can of sparkling water bursting through a splash of water droplets, frozen motion, bright cyan background, high-speed flash photography --ar 3:2 --style raw

a pair of wireless over-ear headphones resting on a floating concrete block, single hard spotlight from above, dark charcoal background, sleek technology advertising, generous empty space on the right --ar 16:9

a skincare jar surrounded by sliced citrus and green leaves on wet dark stone, golden backlight, water droplets, fresh botanical hero image --ar 4:5 --s 200

Raising --s (stylize) pushes the image towards Midjourney's own aesthetic, which can flatter moody hero shots. Lower values keep it closer to your words. Try two or three values on the same prompt before committing.

Lifestyle prompts

Lifestyle shots show the product in use or in a believable setting. The product should still be the clear subject, so describe it first and keep the scene supportive. Shallow depth of field helps separate it from the background.

a ceramic teapot on a wooden kitchen table beside an open book and a linen napkin, morning window light, gentle steam rising, cosy Scandinavian interior, shallow depth of field --ar 3:2

a canvas backpack resting on a rock beside a mountain lake at sunrise, outdoor lifestyle product photography, natural light, 35mm lens --ar 4:5

a hand holding a smartphone in a sunlit café, background softly blurred, candid lifestyle photography, warm natural tones --ar 9:16 --style raw

a scented candle burning on a bathroom shelf next to folded white towels and a eucalyptus sprig, soft evening lamplight, calm spa mood --ar 4:5

a bottle of olive oil on a rustic table with fresh bread, tomatoes and basil, Mediterranean afternoon sun, crisp hard shadows, editorial food styling --ar 3:2

Surfaces, lighting and backgrounds to swap in

Once a structure works, change one ingredient at a time. This keeps a set coherent and shows you which word made the difference.

Surfaces: honed travertine, brushed steel, terrazzo, raw linen, pale oak, wet slate, acrylic riser, sand, moss, rippled water.

Backgrounds: white infinity curve, pastel colour paper, soft gradient backdrop, blurred kitchen, out-of-focus city lights, arched plaster niche.

Lighting: soft diffused studio light, single hard spotlight, rim light, backlight through glass, dappled shadows from leaves, golden hour window light. The lighting guide covers these in depth.

Camera: eye-level, overhead flat lay, low angle for scale, macro detail of texture, 85mm or 100mm for compression.

Colour-matching a brand palette works best when you name the colours plainly ("terracotta and cream") rather than giving hex codes, which Midjourney does not interpret reliably.

Limits: product fidelity and label text

Be realistic about what Midjourney can do. It generates a plausible product, not your product. Proportions, cap shapes, buttons and small details will drift from image to image, and it cannot guarantee an exact match to a real item.

Text is the bigger problem. Midjourney can render short words, but long or exact label copy, ingredient lists, small print and logos come out garbled or subtly wrong. Never rely on it for packaging text you intend to publish.

A workflow that respects those limits:

Use Midjourney for the scene: surface, light, props, mood and composition.

Add --no text, logo, label or describe a "blank unlabelled bottle" so the generated product is clean.

Optionally start the prompt with a photo of your product as an image prompt to steer shape and colour, accepting that it guides rather than copies.

Composite a real photograph of the product, and the real label artwork, into the scene in your editing software.

Match shadows and reflections by hand so the composite reads as one photograph.

Used this way, Midjourney becomes a fast set-builder and moodboard tool, while the product itself stays accurate.

Doing this faster with Dreamdrive

Dreamtune builds prompts like these from dropdowns grouped into Composition, Reference, Technical, Styles, Subjects and Environments, so you can swap surfaces, lighting and camera choices without retyping. My Prompts auto-saves every prompt into project folders, which makes it easy to reuse a winning packshot formula for the next product. Dreamdrive works inside midjourney.com in desktop Chrome.

Q: Can Midjourney make product photos of my actual product?
A: Not exactly. You can use a photo of your product as an image prompt to steer shape and colour, but details drift, so for accurate results generate the scene in Midjourney and composite a real product photo into it.

Q: How do I get a pure white background in Midjourney?
A: Ask for a plain pure white background with soft even studio lighting and an e-commerce packshot style, and add --style raw to reduce artistic interpretation. You may still need to clean the background in an editor for marketplace requirements.

Q: Why is the text on my Midjourney product label wrong?
A: Midjourney is unreliable at rendering long or exact text, so labels, ingredient lists and logos often come out garbled. Generate a blank label and add the real artwork in design software.

Q: What aspect ratio should I use for product photos?
A: Use 1:1 or 4:5 for shop listings and social feeds, 9:16 for stories, and 16:9 or 3:2 for website banners. Check each platform's current image specifications.

---

# Midjourney portrait prompts: formula + 12 examples

URL: https://www.dreamdrive.ai/guides/midjourney-portrait-prompts/
Updated: 2026-10-02

Short answer: A good Midjourney portrait prompt follows this order: framing, subject (age, appearance, expression, clothing), setting, lighting, camera or style cues, then parameters, for example close-up portrait of ..., soft window light, 85mm, natural skin texture --ar 4:5 --style raw. Specific, respectful detail beats long lists of adjectives.

A portrait prompt formula that works

Strong Midjourney portraits come from specific, consistent choices rather than long lists of adjectives. This order covers everything a photographer would decide on set:

Framing: the shot type, such as close-up portrait or medium shot.

Subject: age, appearance, expression and clothing, described plainly and respectfully.

Setting: where they are, or what the background is.

Lighting: one main light with a direction and quality.

Camera and style cues: a lens or medium, and a mood or genre.

Parameters: aspect ratio and any others, always at the end.

Put together, it reads like this:

close-up portrait of a woman in her sixties with silver curly hair and a warm, knowing smile, wearing a rust-coloured knit sweater, in a sunlit kitchen, soft window light from the left, 85mm, shallow depth of field, natural skin texture --ar 4:5 --style raw

For the general principles behind this order, see Midjourney prompt structure.

Lighting for portraits

Lighting decides the mood of a portrait more than any other choice. These setups are reliable starting points; the lighting prompts guide covers many more.

Lighting | Effect on a face | Prompt phrase |

Window light | Soft, natural, gently directional | soft window light from the side |

Overcast daylight | Even and kind to skin, muted colour | overcast daylight, soft shadows |

Golden hour | Warm glow, hair lit from behind | golden hour backlight, warm rim on the hair |

Rembrandt | Moody, classic, sculpted cheekbones | Rembrandt lighting, dark background |

Butterfly | Glamorous, symmetrical, beauty look | butterfly lighting from above |

Split | Half lit, half shadow, intense | split lighting, half the face in shadow |

High-key | Bright, clean, airy | high-key studio lighting, white backdrop |

Neon | Coloured, nocturnal, urban | pink and blue neon light on the face |

Framing and pose

Use film shot types to set how much of the person appears. A close-up focuses on expression; a medium shot adds hands and gesture; a full-body shot shows clothing and stance. The shot types guide lists them all.

Match the aspect ratio. --ar 4:5 or --ar 2:3 suit most portraits. Use --ar 16:9 for cinematic environmental portraits with space around the subject.

Give the eyes a direction. looking directly at the camera feels connected; looking out of the window feels reflective.

Give the hands a job. Holding a cup, adjusting glasses or resting on a table looks more natural than unexplained hands. Hands have improved in recent versions but can still go wrong, so check them and reroll, or use Midjourney's editing tools to redo just the hands.

Choose the angle deliberately. Eye level is honest, slightly high is gentle, low is powerful. See camera angles.

Skin, texture and detail words

Midjourney's default aesthetic tends to smooth skin and polish faces. If you want images that look like real photographs of real people, ask for texture and avoid words that imply retouching.

Words that add realism: natural skin texture, visible pores, freckles, fine lines around the eyes, slight redness on the cheeks, unretouched, film grain.

Words that push towards plastic: flawless skin, perfect face, ultra-smooth, beautiful on its own. Use them only if you want a glossy, stylised result.

Hair and fabric details: flyaway hairs, loose strands catching the light, coarse wool, creased linen. Small imperfections read as real.

Raw mode: adding --style raw and keeping --stylize modest usually helps photographic portraits look less airbrushed.

Diversity and respectful depiction

If you do not describe a person, Midjourney fills in the gaps from patterns in its training data, which can default to narrow ideas of who looks like what. Being specific is both more accurate and more respectful.

Describe people the way you would want to be described. Age, skin tone, hair texture and style, body type and clothing are all fair, neutral details. Avoid loaded or outdated terms.

Do not reach for stereotypes. If a prompt about a profession, culture or nationality returns clichés, describe the actual person and setting instead of the label.

Depict disability, age and religious dress with care. Name assistive devices, hearing aids or headscarves plainly as part of the person's life, not as the subject of the picture.

Avoid caricature. Exaggerated features tied to ethnicity or identity cause harm even when unintended.

Do not create images designed to pass as a real, identifiable person, and check Midjourney's community guidelines for current rules on public figures.

Copy-ready portrait prompts

close-up portrait of a young Black woman with short natural coils and gold hoop earrings, laughing, overcast daylight, natural skin texture, 85mm --ar 4:5 --style raw

environmental portrait of an elderly Japanese potter at his wheel, clay on his hands, soft window light, quiet workshop, 35mm documentary photograph --ar 3:2 --style raw

medium close-up of a red-haired teenager with freckles, looking out of a rainy bus window, cool blue tones, film grain --ar 4:5

studio portrait of a bearded man in his forties in a navy turtleneck, Rembrandt lighting, dark grey backdrop, calm confident expression --ar 4:5 --style raw

beauty portrait of a South Asian woman with dark wavy hair, butterfly lighting from above, soft glowing skin with visible pores, pale pink backdrop --ar 4:5

golden hour portrait of a father carrying his daughter on his shoulders in a wheat field, backlit hair, warm haze, 85mm --ar 2:3

portrait of a wheelchair basketball player resting on the court, sweat on her brow, rim light, dark arena background, determined gaze --ar 4:5 --style raw

close-up of a woman in her seventies wearing a patterned headscarf, fine lines around her eyes, gentle smile, warm afternoon window light --ar 4:5 --style raw

cinematic portrait of a man in a rain-soaked jacket under pink and blue neon light, night city reflections, shallow depth of field --ar 16:9

high-key headshot of a smiling nurse in teal scrubs, clean white background, soft even light, friendly and professional --ar 1:1 --style raw

black and white portrait of a jazz trumpeter mid-note, low-key lighting, smoke in the air, high-contrast film grain --ar 2:3

full-body portrait of a non-binary person in a tailored oversized suit, standing in a concrete stairwell, overcast light, fashion editorial --ar 2:3 --style raw

Treat these as starting points: swap the subject details for your own, keep the structure, and adjust one element at a time so you can see what each change does.

Doing this faster with Dreamdrive

Dreamtune builds prompts from dropdowns grouped into Composition, Reference, Technical, Styles, Subjects and Environments, with Simple or Complex output. My Prompts auto-saves every portrait prompt into searchable project folders, and Style Vault keeps the --sref codes you like with preview images.

Q: How do I make Midjourney portraits look realistic?
A: Ask for natural skin texture, visible pores or fine lines, use a real lighting setup such as window light, and add --style raw. Avoid words like flawless or perfect skin, which push towards a plastic look.

Q: What aspect ratio is best for portraits?
A: 4:5 and 2:3 suit most portraits. Use 1:1 for headshots and profile pictures, and 16:9 for cinematic portraits with space around the subject.

Q: How do I avoid stereotypes in AI portraits?
A: Describe the actual person with neutral details such as age, skin tone, hair and clothing rather than relying on a label like a nationality or profession. If results drift into cliché, add more specific detail.

Q: Why do my Midjourney portraits all look the same?
A: Vague prompts let the model fall back on its defaults. Specify age, features, expression, lighting and setting, and try a little --chaos to see more variety in the grid.

Q: How do I fix bad hands in a portrait?
A: Give the hands a clear task, such as holding a cup, then reroll or use Midjourney's editing tools to redo just the hands. Hands are better in recent versions but still worth checking.

---

# How to Make Consistent Characters in Midjourney

URL: https://www.dreamdrive.ai/guides/midjourney-consistent-characters/
Updated: 2026-10-02

Short answer: To keep a character consistent in Midjourney, reuse one fixed, detailed description, anchor it with reference images of the character, lock the look with a style reference, and keep everything in a character sheet you copy from every time. Midjourney has also offered dedicated character or omni reference features in some versions, so check the current docs for what your model supports.

Why characters drift

Midjourney does not remember people between prompts. Each image is generated fresh from your words, parameters and any references, so small changes in wording, angle, outfit or lighting can produce someone who looks like a cousin rather than the same person. Faces drift most, followed by hair, proportions and clothing details.

No single trick solves this. Consistency comes from stacking several controls so there is less room for the model to improvise. The techniques below work across recent versions and can be combined.

Technique 1: a fixed, detailed description

Write the character once, in a specific order, and paste it unchanged into every prompt. Vague words like "a young woman" invite a different person every time. Distinctive, concrete details give the model something to hold on to.

Identity basics: approximate age, build, skin tone.

Face: face shape, eye colour, eyebrows, nose, freckles, scars.

Hair: colour, length, texture, a signature style.

Signature clothing: one or two items that always appear, such as a mustard raincoat or round brass glasses.

Rendering style: the medium and look, which should not change between images.

Example character block:

Mara, a woman in her early thirties, narrow face, light brown skin, dark green eyes, thick straight eyebrows, short asymmetric black bob, small scar through her left eyebrow, mustard yellow raincoat, round brass glasses

Then vary only the scene around it:

Mara, a woman in her early thirties, narrow face, light brown skin, dark green eyes, thick straight eyebrows, short asymmetric black bob, small scar through her left eyebrow, mustard yellow raincoat, round brass glasses, waiting at a rainy tram stop at night, cinematic still --ar 16:9 --style raw

Mara, a woman in her early thirties, narrow face, light brown skin, dark green eyes, thick straight eyebrows, short asymmetric black bob, small scar through her left eyebrow, mustard yellow raincoat, round brass glasses, reading in a crowded library, warm lamp light, cinematic still --ar 16:9 --style raw

Keeping the character block at the start of the prompt gives it more weight than the scene.

Technique 2: image prompts of your character

Once you have one image of the character you like, use it as an image prompt. Place its URL at the start of later prompts, followed by your character block and the new scene. Raise --iw if the character is drifting and lower it if every image starts copying the original pose and background.

Tips for reference images:

Pick a clear, well-lit image where the face is visible and unobstructed.

Crop to the character so the background does not dominate.

Build a small set: a front view, a three-quarter view and a full-body shot.

Upscale your best images so you have clean references to reuse.

Image prompts carry composition as well as identity, so expect some of the original framing to leak through. That is the trade-off.

Technique 3: lock the look with style references and seeds

A character can be the same person yet feel inconsistent because the rendering changes: one image is painterly, the next glossy. A style reference fixes the visual language so your character always lives in the same world. Use the same --sref image or code, at the same --sw, across the whole series.

A --seed can also help. Reusing a seed with a near-identical prompt gives a similar starting point, which is useful for small changes like a different expression. It will not hold a character across very different scenes on its own, so treat it as a fine-tuning tool rather than the main method. See our seeds guide for how they behave.

Control | What it keeps consistent | Limits |

Fixed description | Key features and signature clothing | Faces still vary in detail |

Image prompt + --iw | Face, hair, overall likeness | Also pulls in pose and composition |

--sref + --sw | Rendering style, palette, mood | Does not control identity |

--seed | Starting noise for near-identical prompts | Weak across different scenes |

Character or omni reference (some versions) | Identity across new scenes | Availability and syntax depend on version |

Vary (Subtle) and editing | Small changes to an existing image | Not for new scenes |

Dedicated character reference features

In some versions Midjourney has offered a dedicated character reference, and later an omni reference, designed specifically to carry a person or object from a reference image into new images. When available, these are usually the strongest single tool for consistency, especially for faces.

Because these features have changed between model versions, including how they are called and which versions support them, check Midjourney's current documentation before relying on them. Two practical notes apply whatever the version:

They work best with a clean reference of the character, so the reference set you built in technique 2 is still worth having.

They combine with everything else here. A dedicated reference, a fixed description and a fixed style reference together beat any one of them alone.

Keep a character sheet

Professionals keep a character sheet, and you should too. It turns consistency from memory into a copy-and-paste routine. Store, for each character:

The exact character block text.

Links to your best reference images: front, three-quarter, full body.

The style reference code or URL and the --sw value.

Any seed that produced a good base image.

Parameters you always use, such as --ar, --style raw or --s.

A "do not change" list: features that must appear in every image.

You can also ask Midjourney for a turnaround to populate the sheet:

character turnaround sheet of Mara, a woman in her early thirties, short asymmetric black bob, mustard yellow raincoat, round brass glasses, front view, side view and back view, plain light grey background, clean concept art --ar 3:2

Expect to curate. Generate several grids, keep the images that match, and discard the rest. Consistency is partly selection.

Doing this faster with Dreamdrive

My Prompts auto-saves every prompt into project folders, so a character's working prompts stay together, searchable and one click from reuse. Style Vault stores the --sref codes you use for a series with preview images, so the look stays the same from scene to scene.

Q: Can Midjourney keep the same character across images?
A: Yes, with effort. Combine a fixed detailed description, image prompts of your character, a consistent style reference and, where your version supports it, a dedicated character or omni reference feature.

Q: Does using the same seed keep a character consistent?
A: Only partly. A seed gives a similar starting point for near-identical prompts, but it will not hold a character across very different scenes on its own.

Q: Does --sref keep a character's face consistent?
A: No. A style reference transfers the look, such as palette and rendering, not identity. Use it to keep the visual world consistent and other methods to keep the person.

Q: What is a character sheet in Midjourney?
A: It is your own record of a character's exact description, reference images, style code, seeds and parameters, which you copy from for every new prompt.

Q: Does Midjourney have a character reference parameter?
A: Midjourney has offered character reference and omni reference features in some versions. Availability and syntax depend on the model version, so check the current documentation.