Library

Drag. Drop. Done.

The uploader is a real pipeline, not a file input. Multi-file drag & drop with progress and duplicate detection — and by the time you land in the editor, posters, filmstrips, proxies, and a transcript are already waiting.

What happens after the drop.

Every upload runs the full ingest pipeline automatically. You never wait for a spinner — the library is working while you do something else.

01
Drop

Multi-file drag & drop at /upload — any logged-in user. Progress per file; duplicates are flagged by content hash before they land.

02
Probe & fingerprint

ffprobe pulls metadata; sha256 fingerprints the asset — duplicates get caught, merged, and never pile up twice.

03
Proxies + filmstrip

720p and 540p editorial proxies plus a scrub rail render automatically. Editing stays instant on laptop Wi-Fi — masters never leave storage.

04
Transcribe & tag

Word-level transcript for search and captions; auto-tagging files it into collections. Everything is searchable the moment it lands.

Big files, bad Wi-Fi, no drama

Uploads that survive real life.

  • Direct multipart, 512 MB per file

    The everyday path — drag, watch the progress bar, get to work.

  • Growing upload sessions

    Chunked, resumable sessions via the API: append chunks, check status, finalize when ready. Big files and unstable connections stop being a gamble.

  • Watch folders

    Drop files into an S3 watch/ prefix; every 60 seconds the pipeline ingests them — proxies, transcript, tags — and pings you when they're ready.

  • Duplicate detection at the gate

    sha256 grouping and a merge path in the library — the same take never enters twice.

  • First-run assist

    New to the editor? A first sequence already exists, and your first action — even a drop — starts a project and applies.

What you get, plain.

  • Dedicated uploader page — multi-file, progress, duplicate detection, any logged-in user
  • 512 MB per file direct multipart; resumable growing sessions via API for the rest
  • Watch folders on S3 prefixes — ingest every 60 seconds, fully processed
  • Proxy ladder on every asset — 720p + 540p + scrub rail
  • Word-level transcript on every upload — searchable, caption-ready
  • Duplicate detection with sha256 grouping and merge
  • The screen recorder (/record) runs on the same chunked plumbing — a dropped connection never kills a take

"Storage is cheap. Re-uploading footage because a connection dropped is not."

— growing upload sessions

Your footage is a
drag away.

Proxies, transcripts, and tags before you even open the editor.