Screenshot → recreation

Turn any image you can screenshot — a look you admire, a competitor frame, an old asset with no surviving prompt — into an editable generation of your own.

The recipe

1. Paste it — copy the image and press ⌘V on the canvas. An Image input node appears with it attached.
2. Reverse the prompt — wire it into an LLM node’s image input and ask for a reverse-engineered generation prompt: subject, composition, lens, lighting, grade, mood.
3. Regenerate — wire the LLM’s text output into Image Gen (or Video Gen to bring it to life) and run.
4. Remix — now it’s text: edit the recovered prompt to change the subject while keeping the look, or bolt on a library style to pull it on-brand.

Tips

• Ask the LLM for “a single-paragraph generation prompt” — bulleted analyses generate worse than flowing descriptions.
• For a closer match, also wire the original screenshot into one of Image Gen’s reference slots so the model sees it alongside the recovered prompt.
• The recovered prompt is a great starting point for a Collection brief — one screenshot becomes a whole consistent series.