Describe what you want the AI to create.
Claude refines your prompt using the image as context, then your image model renders the pixels into the masked area.
Remove & Expand run locally (no prompt). Fill & Add use the Claude API for prompt intelligence + your image model.
Describe what you want the AI to create.
functions/api/claude.js and set the ANTHROPIC_API_KEY secret in the dashboard. The key never reaches the browser. See DEPLOY.md.localStorage) and exposes it to anyone using the page — use only for local testing.AIza… key (or deploy functions/api/gemini.js with a GEMINI_API_KEY secret for proxy mode). Model id default gemini-3.1-flash-image. Paid per image (~$0.04).--api --cors-allow-origins=*, URL http://127.0.0.1:7860. See LOCAL_SD.md. ⚠ Run Photo Ray over http://localhost (not the https site) for local models — mixed-content block.Extend the canvas by the amounts below. New pixels are generated locally from the surrounding image (no prompt needed).