Extract a style
Extraction uses AI to derive style blocks from sources you provide — text, images, or documents. You review the result and apply the blocks you want to your style board. This guide covers the sources, the flow, and what your plan changes.
What you can extract from
| Source | Details | Availability |
|---|---|---|
| Text | A description, brand copy, or a writing sample — up to 100K characters. Naming a reference aesthetic works too (“Bauhaus”, “1970s paperback cover”). | Everyone |
| Images | Drag-and-drop or upload, up to 20 MB each (compressed automatically). | Everyone — up to 2 images per style (Pro: 10) |
| PDFs & documents | A brand guideline or style document, up to 4 MB. | Pro only |
You can combine sources in one run — text plus images, for example — and StyleRef fuses them into one extraction.
Run an extraction
Open the Extract Style box
On the style board, the Extract Style box is the first card. Paste or type your text, attach files, and select Extract Style. The button shows the credit cost before you run it.
Wait for the analysis
Extraction runs in the background — usually well under a minute, longer for large inputs.
Review the extracted blocks
The preview shows every extracted block, grouped by section. For each block, choose Add (keep what’s on your board and add this) or Replace (overwrite the board’s block of the same type).
Apply
Select Apply N items to merge your selected blocks into the style board. From there, refine them like any other block.
Plans, credits, and AI tiers
- Anonymous — one free extraction, no account needed. Sign up to keep going (your board carries over).
- Free — extractions spend credits; the cost is shown on the Extract button. Extraction runs on Lite AI — a quick style read.
- Pro — unlimited extraction on Advanced AI: a deeper analysis of your sources that produces a more complete extraction, plus PDF/document sources and more files per run.
If an extraction fails, the credits it spent are automatically refunded — you’re only charged for successful runs. If a run keeps failing, try a clearer source: a sharper image or a more specific description.
Extracting outside the web app
Extraction is available in the web app, over MCP (extract_style), through the REST API, and from the CLI. MCP and REST return the blocks to the caller (or save a draft) instead of the on-board preview.
The ComfyUI connector deliberately does not extract: its nodes load and apply styles you already have. To turn an image into a style, use the web app — or the API/MCP/CLI if you are scripting it — then load the result in ComfyUI.
Already extracted something once? Read it back instead of paying again: the web app keeps your extraction history, and GET /api/v1/me/extractions exposes the same history to scripts and agents.
Next steps
- Create a style — refine the applied blocks on the board.
- Plans & limits — the credit system in full.
- Use your style in AI tools — put the result to work.