Turn your Stash library into PMVs

If you run Stash, you've already done the boring part. Scenes tagged, performers identified, everything catalogued — often over years. That metadata is worth more than people realise, because it can drive editing directly instead of just sitting in a library.

Onset Engine can read a local Stash instance and pull that curation in. This walks through it.

What actually gets imported

Onset matches your ingested source files to Stash scenes by file path, then imports:

• Performers — each becomes a "Subject" in Onset that you can filter by and reference in drivers
• Tags — optional, off by default (Stash tag lists get long, and not all of it is useful for editing)

The part that makes this more than a metadata copy: each imported performer also gets an AI fingerprint built from the clips it matched. That means Onset can then find more clips of that performer — including from footage that was never in Stash, or was never tagged.

Before you start

• Stash running locally (the default http://localhost:9999 is what Onset assumes)
• Your footage already ingested into Onset — the import matches against files Onset knows about
• An API key only if your Stash requires login; most local setups don't

The paths have to line up. If Stash sees a file at E:\scenes\clip.mp4, Onset needs to have ingested that same file at that same path. If you keep everything on one drive and ingest from where it lives, this just works.

Doing the import

1. Open the SUBJECTS tab in the asset bin
2. Click ⬇ STASH
3. Leave the URL as http://localhost:9999 unless yours differs
4. Hit Preview (Dry Run) first — it reports how many of your sources matched and which performers it found, without writing anything
5. If the numbers look right, hit Import

Or from the command line:

OnsetEngine.exe stash-import --dry-run
OnsetEngine.exe stash-import

Add --include-tags if you want Stash tags as well as performers.

Always dry-run first. If it reports far fewer matched sources than expected, it's almost always a path mismatch — Stash and Onset are looking at the same files by different routes (mapped drive vs UNC path, that sort of thing).

What to do with it

Once performers are in, a few things open up:

Build an edit around one performer

In a driver, reference the performer by name in a tier's tags:

"tags": ["@PerformerName"]

Put that on the peak tier and the drops will pull from that performer specifically. Put it on every tier and the whole edit is theirs. The driver pack is a good starting point to modify.

Find clips Stash doesn't know about

Hit 🔍 MORE on a Subject card. Onset re-scans your whole library for clips resembling the ones it already has, and hands you a review queue — thumbnails, similarity scores, and a preview button that jumps to the exact segment. Approve the good ones, ignore the rest.

This is the bit that compounds: every approval sharpens the fingerprint, so the next search is better than the last. Footage you never got round to tagging in Stash gets pulled in anyway.

Filter the bin

Hit FILTER on any Subject to scope the asset bin to just that performer — useful for hand-picking before a render.

Worth knowing

• Matching is per source video, not per clip. Stash metadata is scene-level, so every clip cut from a matched scene gets the performer tag — including moments where they're not on screen. The review queue and Find More are how you tighten that up.
• Re-running is safe. Import again after tagging more in Stash; existing subjects update rather than duplicating.
• Nothing leaves your machine. It's your metadata, read over localhost, into a local database.
• No face recognition involved. The fingerprint is general visual similarity, not biometric identification — which also means it happily works on things that aren't people.