Feature
Drop files in a bucket.
They appear, processed.
Ingest shouldn't be a chore you perform. Point a watch folder at an S3 prefix, and anything that appears there flows through the full pipeline on its own — storage, proxies, transcript, auto-tags — and shows up in the library as a notification.
The easy explanation
Like a mailbox that does the work.
Think of a watch folder as a mailbox with a processing plant behind it. You (or a camera, or a producer in the field) drop files into the S3 watch/ prefix. Aperture notices, and each file gets: secure storage, fast proxies for playback, a transcript if it has audio, and AI tags. You never touch the pipeline — it just works.
Resumable browser uploads
Big files upload in chunks. A dropped connection resumes — it never costs you a take.
Duplicate check at the gate
The sha256 fingerprint is taken before anything enters the library.
Notifications on arrival
The team learns "3 new assets" — not "did the upload finish?".
Everything ingest does, automatically.
- Secure storage on S3/MinIO — your library is the tenant library, shared
- Playback proxies generated for smooth browser scrubbing
- Whisper transcription — searchable the moment it lands
- AI auto-tagging with description (admin toggle)
- sha256 fingerprint — duplicates blocked, not discovered
- 512 MB import cap with SSRF host allow-listing for stock
"The library used to start when you opened the app. Now it starts when the file exists."
— watch foldersThe file did nothing wrong
It just sat there. Then it edited itself.
Drop files into the bucket and walk away. Storage, proxies, transcript, tags — processed while nobody watches. The library starts where the file exists.
Watch the folderDrop files in a bucket.
They appear, processed.
Watch folders feed the library while nobody watches. Uploads survive bad connections.