Nomad Studio/ Documentation/ Create/ Layout Reference

Create · Layout composition

Layout Reference

Layout Reference works like Layout Assist, but adds an optional reference image — a character, a background, a style, or a general scene. Vision analyzes it and blends the extracted details into the composition. For a character reference specifically, the AI extracts a complete verbatim appearance (skin tone, hair, makeup, every garment, every accessory) and reproduces it in full inside that element's description, then adds your own scene and pose direction on top — so the same character can be redrawn in a new scene while keeping their exact look.

All Layout composition flows share one coordinate system: a normalized 1000×1000 canvas, with each element's bounding box written as [y_min, x_min, y_max, x_max] (origin top-left, y increases downward). The LLM never draws pixels directly — it writes this JSON layout, and the Composition step renders and lets you fine-tune it before generating.

At Scene and Master depth, the model is instructed to work through a 7-point checklist on the reference photo before writing any JSON: camera vantage point, apparent focal length, subject orientation and gaze, foreground anchors, the background's focal point, medium/colour grading, and environment-specific colours as real hex values — all of it used to inform the final layout with photographic precision.

Guided walkthrough

1. Step type: Workflow picker

   Choose engine

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

2. Step type: Prompt input

   Composition idea

   Fragments separated by ;, same as Layout Assist.

3. Step type: Image picker (optional)

   Reference image

   Has vision analysis. Choose what it represents: character, background, style, or scene (default character).

4. Step type: Parameter group

   Format

   Aspect ratio, seed, steps, CFG, image count — open by default.

5. Step type: LLM transform — llm.ideogram_compose

   Compose

   Blends your fragments with the reference analysis into a full layout, at the chosen depth (Quick → Master).

6. Step type: Ideogram composer

   Composition

   Review and fine-tune before generating.

7. Step type: Action — generation.submit_queue

   Generate Image

   Sends the composed layout to the queue.

8. Step type: Result view

   Result

   Save, Upscale, Send to Refine, Regenerate.

Examples