Artlist AI Agent, Flows and Style Kit Explained

Single prompts are fine for single images. Real projects involve a sequence of steps, a look that must stay on brand and characters that must stay recognisable, and that is where Artlist's newer tools come in. They sit inside Artlist AI Tools alongside the image, video, voiceover and music generators and share the same credits. This guide explains what each one does, according to Artlist's help centre, and when it is worth using.

The AI Agent: generate by conversation

The Agent replaces menus with dialogue: describe what you want, get an image or video, and refine it with follow-up messages or reference files. Open Image or Video in the side navigation and click the speech-bubble icon in the prompt area. Session starters include "Help me write a prompt", "Create a video", "Describe an image", "Create an image" and "Edit an image".

• Free to talk, paid to generate. Chatting, refining prompts and visualising references cost no credits; generating or refining an image or video uses credits at the standard rate. Generations over 1,000 credits ask you to confirm first.
• Prompt help without generating. Say "Help me improve this prompt…" and the Agent works on the wording only.
• Context helps. Tell it what the work is for ("a presentation for my co-workers about the financial crisis") and it writes a prompt to suit.
• Image first, then video. Artlist describes the optimal video flow as creating an image, then animating it.
• Memory lasts one session. Start a new session and the Agent no longer knows the earlier conversation, so keep your own notes.

The Agent also explains settings, suggests models when asked and moves between formats in one session: text to image, image to video, audio to video.

Flows: a pipeline on a canvas

Flows is a visual, node-based canvas for building multi-step generation pipelines. Instead of generating an image, downloading it and uploading it somewhere else to animate, you connect the steps once and run them together. Open Flows from the side navigation and start from + New Flow or a template.

Ports are colour-coded by type (text, image, video, audio), so only compatible nodes connect. Every run keeps its result in the node's gallery rather than overwriting it, and you choose which result feeds the next node. Each generator shows its credit cost before you run it, a failed node shows its own error and Retry button, and the canvas saves automatically. Flows suits work you repeat: a product shot to animated clip, or one prompt tested on several models side by side.

Style Kit: a brand look in one click

Style Kit stores a brand's look once and applies it to generations. A kit holds up to two reference images, an optional colour palette and optional guidelines text describing tone and direction. Create one from the Style Kit selector in the prompt panel, then select it before generating.

• Your own reference images take priority; kit images fill the remaining reference slots.
• The kit's guidelines are applied behind the scenes, so your visible prompt stays clean.
• Only one kit can be active per generation.
• Image generation is supported; video support depends on the model.
• Kits also work in Artlist Studio and through the Artlist MCP connector.

Write guidelines as you would brief an illustrator: palette, lighting, texture, what never to show. Keep the two images genuinely representative; one weak reference pulls every generation off-brand.

References: characters, locations and props

References save a Character, Location or Prop so you stop re-uploading the same image. Create one from the + icon in the prompt box, choose My References, then New: pick a type, upload a photo (or, for a Character, a multi-angle Character Sheet), name it and click Add. Attach it from My References and type @ to tag it by name in the prompt. You can also turn any creation into a Reference with Use, then Save as Reference. Artlist recommends a clear, front-facing, well-lit photo, and lists Veo 3.1, Gemini Omni Flash, the Seedance 2.0 family, Seedream 5.0 and the Nano Banana image models among supported models at the time of writing.

Putting them together

1. Create a Style Kit for the client and References for the recurring cast and locations.
2. Use the Agent to turn the brief into prompts, without spending credits.
3. Build a Flow: prompt, image generator with the kit, video generator using the chosen image as the start frame.
4. Run node by node while testing, then run the full flow for each shot.
5. Record the model and settings of every approved result outside the session; the Agent's memory ends with it.

For the generator settings themselves, see the Artlist AI video generator guide.

Try it

Help me improve this prompt for a 6-second product video: a matte green water bottle on a stone ledge by a window, morning light, a slow push in, birdsong outside. Keep it on brand with my Style Kit and suggest a model with good audio.

Send it to the Agent as a prompt-help request, so no credits are used. Take the improved prompt into a Flow with an image node and a video node, and compare two models side by side before committing to a final render.