NightCafe Generation Guide: Prompts, Models and Credits

NightCafe (creator.nightcafe.studio) is one of the easier places to start with AI images. Dreamdrive saves every prompt you run automatically, with its picture, model and settings, into the project you choose, so you can search your Library for it later, copy it with Copy prompt or drag the picture into another tool. This guide covers how to choose a model, how to write a prompt that works there, when to leave preset mode, and how to stretch your credits.

Choosing a model

NightCafe does not run a single model. In the creator you choose from a list that has included Stable Diffusion variants, SDXL-based models, Flux models, DALL-E and others at various times. The list changes as models are added and retired, so treat what follows as guidance about types rather than a fixed menu.

• Flux-type models read natural sentences well, handle text in images better than older models and generally need less prompt engineering. A good default for photoreal and illustration work.
• SDXL-type models respond to comma-separated phrases, support a negative prompt in advanced mode and have a huge vocabulary of styles. Good when you want fine control.
• Older Stable Diffusion models cost fewer credits and are fine for quick ideas, but struggle with hands, text and complex scenes.
• Models from other providers, where offered, bring their own strengths and sometimes their own content rules.

Credit costs differ by model and by size, and the interface shows the cost before you confirm. Dreamdrive saves every prompt you run automatically, with its picture, model and settings, into the project you choose, so you can search your Library for it later, copy it with Copy prompt or drag the picture into another tool.

Writing the prompt

A clear prompt has the same bones on every model. State what is in the picture, where it is, how it is lit and what it should look like as an image.

1. Subject and action. "A heron standing in shallow water, catching a fish."
2. Setting. "A misty reed bed at sunrise."
3. Light and atmosphere. "Low golden light, mist catching the sun."
4. Image type and style. "Wildlife photograph, telephoto lens, shallow depth of field." Or "gouache painting, loose brushwork".

On Flux-type models write this as one or two sentences. On SDXL-type models the same content as comma-separated phrases works well. Avoid long strings of quality words; "masterpiece, trending, 8k" adds little on current models and pushes everything towards the same glossy finish. Do not prompt for brand logos or real people.

Preset mode versus advanced mode

NightCafe’s creator has a simple flow where you type a prompt and pick a style preset, and an advanced flow where more controls appear. The preset applies a style description and sensible settings behind the scenes, which is why presets produce reliable results for beginners.

Stay in preset mode while you learn what each style does. Move to advanced mode when you need any of these: a negative prompt, a specific aspect ratio, a fixed seed to reproduce or vary an image, control over how closely the model follows the prompt, or a chosen number of outputs. In the current interface the labels for these controls vary by model; the aspect ratio and seed are the two worth learning first, because they are the most useful for working to a brief.

A preset is just words and settings. If you like what a preset does but want to edit it, generate with it once, open the result’s details to see the full prompt it produced, and copy that text into advanced mode as your starting point.

Settings at a glance

For a deeper look at presets and the advanced controls, see NightCafe styles and settings.

Making credits go further

• Draft small, finish large. Explore with one image per run at the default size. Pay for a larger size only on a prompt you have already proven.
• Use the cheaper models to test composition, then move the same prompt to a Flux-type model for the final.
• Fix the seed once you have a keeper and change one word at a time. Random seeds while refining waste credits on chance.
• Join the daily challenge if you enjoy it. Community challenges are part of the site and a low-pressure way to practise prompts on a theme.

Keeping your prompts and settings

Every creation on NightCafe keeps its prompt, model, preset and settings on your account page, and you can duplicate or evolve a creation from there. Downloaded images do not reliably carry that data, so when an image matters, copy the prompt, model, preset, aspect ratio and seed into your own notes. If you work across several platforms, see tracking prompts across AI tools.

Try it

Model: a Flux-type model
Prompt: A heron standing in shallow water at sunrise, catching a small fish, misty reed bed behind it, low golden light catching the mist. Wildlife photograph, telephoto lens, shallow depth of field.
Preset: none. Aspect ratio: 3:2. Number of images: 1.

Run it once, then run it again with a painterly preset applied. Comparing the two shows you exactly what the preset adds, which is the quickest way to learn what each preset is for.