Gallery
Everything you have made, newest first, grouped by day.
Click one and its full record appears: the prompt, the model, the seed, the steps, the guidance — everything needed to make it again. Filter by model, by adapter, or by date, and search your own prompts.
Thumbs up or down on a render, or put two side by side and say which is better.
Both need History logging switched on (System → Settings). It is off by default, and while it is off nothing about your renders is recorded, so there is nothing to attach a rating to — the app will say so rather than pretending. Renders made after you turn it on can be rated.
Ratings feed the Insights panel on the Agents screen, which reports what has actually worked: which models you rate highest, which step counts and guidance values, and the words that keep appearing in your best prompts. Pairwise comparisons are recorded but do not yet feed those figures.
Renders are ordinary PNGs in your outputs folder, with a .json sidecar
carrying the same settings. Deleting from the Gallery deletes the file.