Preview · Stash plugin
Rediscover scenes in your Stash library.
Curator uses your library metadata, viewing history, and feedback to recommend scenes you already have. Open Why this? to see why a scene appeared and correct the model when it gets something wrong. Connect StashDB for optional external metadata leads scored against your local taste model.
Install
Install and build your first recommendations
Requires Stash v0.31 and a python executable for the plugin launcher. Curator ships compiled core binaries for supported platforms; StashDB is optional. Add this URL under Settings → Plugins → Available Plugins:
https://mrx-31415.github.io/stash-curator/index.ymlInstall Stash Curator, reload plugins, open the compass, and select Sync and build recommendations in the setup checklist or under Manage → Tasks. The first build can take several minutes. Check prerequisites and follow the setup guide →
When it is ready, open Recommendations → For You. Try Why this? on a suggestion, then give a thumbs up/down or answer Curate → Tag questions. Feedback affects rankings after a model update.
Curator in action
See scenes move before you choose
Recommendations, Find, and Pair picks move through this eleven-second silent tour. Use the video controls to pause or replay it.
The video opens on For You recommendations with six moving scene previews, switches to Find's Similar results, then shows two scenes in Pair picks and advances to another pair after a choice.
Demo excerpts from Big Buck Bunny and Glass Half (CC BY 3.0), and Caminandes: Llamigos and The Daily Dweebs (CC BY 4.0). © Blender Foundation. Excerpts shortened, muted, and cropped.
Recommendations →
Start with For You, find strong unwatched matches in Best Bets, or return to familiar scenes in Revisit. Stretch, Blind Spots, and Dormant help you explore more of your library, and New Arrivals picks from what you just added.
Find →
Find opens on Expand, which ranks a refreshed pool of StashDB leads when connected. Similar finds related scenes in your library or on StashDB; Performer Hunt follows a performer’s catalog. External results are leads, not proof that a scene is available locally.
Curate →
Answer Tag questions or rate tags in Tag sentiment to teach Curator what you like. After a model update, Impact shows which preferences and recommendations changed.
Privacy and safety
Runs locally. Your history, feedback, and model stay in a SQLite sidecar you control; StashDB is optional, read-only, and never sees your model. The only Stash mutation is an explicit, reversible Prune tag — Curator never deletes media. See Privacy.
Preview status
Curator targets Stash v0.31. It remains preview software and pre-1.0. The first sync/model build can take several minutes on a large library. Its persistent background tasks can keep recommendations up to date, and you can schedule library syncs and backups. External discovery needs a configured StashDB connection. Start with Getting started, then read Using Curator.
Acknowledgements and project provenance
The idea was inspired by Restash by Espionage9248.
Stash Curator is primarily generated with AI coding agents under human direction, review, and testing.