Nomad Studio/ Documentation/ Create/ Describe Image

Create · Assisted

Describe Image

Describe Image starts from a picture instead of words. Upload a reference, the local vision model writes a description of what it sees, and you can optionally say what you'd like changed — a second LLM pass applies just that instruction to the description before it becomes your final prompt.

Reach for it when you have a reference image and want a prompt that starts from what's actually in it, rather than typing a description yourself.

Guided walkthrough

1. Step type: Workflow picker

   Choose engine

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

2. Step type: Image picker

   Reference image

   Required. The image you want described.

3. Step type: Vision describe

   Describe reference

   Runs the local vision model over the uploaded image and writes a description of it.

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

   Description

   The vision output lands here, fully editable. Placeholder text reminds you to click "Describe reference" above if you haven't yet.

5. Step type: Prompt input

   Your idea (optional)

   What you'd like to change about the description — leave empty to generate the description exactly as Vision wrote it.

6. Step type: LLM transform — llm.edit_prompt

   Apply idea to description

   Takes your change instruction and the current description as context, and rewrites the description to apply just that change.

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

   Final prompt

   The edited result, still editable, prefilled from the previous step's output.

8. Step type: Parameter group

   Parameters

   Seed, steps, CFG, image count.

9. Step type: Action — generation.submit_queue

   Generate Image

   Sends the final prompt (or the plain description, if you skipped the idea step) to the queue.

10. Step type: Result view

    Result

    Save, Upscale, Send to Refine, Regenerate, Re-enhance.

Examples