Nomad Studio/ Documentation/ Create/ Chat CivitAI

Create · Conversational

Chat CivitAI

Chat CivitAI is Chat's style-locked sibling: you paste a CivitAI generation block once at the start, and its visual style — characters, clothing, colors, lighting, composition — stays fixed as an anchor through every turn of the conversation. Only the content you explicitly ask to change actually changes.

The CivitAI reference can be cleaned up before use (strip out its described elements, characters, or background if you only want the pure style), and there's a one-click "Generate Scene Prompt" shortcut right after setting the scene, so you don't have to wait for the first full loop turn to see a result.

Guided walkthrough

1. Step type: Workflow picker

   Choose engine

   Pick the imported engine, or auto-selected if you only have one.

2. Step type: CivitAI paste

   CivitAI style reference

   Required. Comes with optional strip buttons to remove elements, characters, or background from the parsed block before it's used as the anchor.

3. Step type: Prompt input

   Set the scene

   Describe the initial scene. Confirming it stores the world context for the session. A "Generate Scene Prompt" quick action right here runs a first LLM pass combining your scene with the CivitAI style in one go, without waiting for the main loop below.

   Show the exact system prompt for the quick action

   You are an image prompt generator for AI image generation. You receive a visual scene description and a CivitAI style reference, and you must write a single comprehensive image prompt.
   Rules:
   - The prompt must be fully self-contained and detailed enough to generate the image on its own.
   - Use the CivitAI style reference as the visual and technical anchor: characters, clothing, colors, lighting style, composition.
   - Apply the scene description as the subject and content direction.
   - Write in the comma-separated descriptive style used by Stable Diffusion and Flux image models.
   - Output ONLY the image prompt text. No JSON, no labels, no explanation.

4. Step type: LLM transform — llm.story_continue

   Generate scene prompt

   The main loop step. On later turns, takes your change instruction plus the full chat history and the CivitAI style, and writes the next prompt — preserving character, setting and composition style unless told otherwise.

   Show the exact system prompt

   You are an image prompt generator for AI image generation. You receive a visual scene description and a chat history of previous scenes, and you must write the next single image prompt.
   Rules:
   - The prompt must be fully self-contained and detailed enough to generate the image on its own.
   - Preserve the same characters, setting, and composition style across iterations unless the instruction says otherwise.
   - Only change what the user's instruction asks for — keep everything else stable.
   - Write in the comma-separated descriptive style used by Stable Diffusion and Flux image models.
   - Output ONLY the image prompt text. No JSON, no labels, no explanation.

5. Step type: Prompt input (pre-filled)

   Scene prompt

   Editable before generating, prefilled by the step above.

6. Step type: Parameter group

   Parameters

   Seed, steps, CFG, image count.

7. Step type: Action — generation.submit_queue

   Generate Image

   Sends the scene prompt to the queue; on success it's appended to the session history and the scene counter advances.

8. Step type: Result view

   Result

   Same timeline view and five actions as Chat: Save, Upscale, Regenerate, Continue story, End story.

9. Step type: Prompt input (loop trigger)

   Change the scene

   Appears after "Continue story". Spatial, temporal or narrative change — the style stays fixed. Loops back to "Generate scene prompt".

Examples