v1.1.6 — A Rebuilt Download Queue & Local Audio Cleanup
A rebuilt download queue
- New queue page: overall progress in the tab title, per-task state, and pause, retry and cancel in one place.
- Preview the merged result — audio and video together — before saving anything.
- Files are written straight into the folder you choose, with the space a batch needs shown before it starts.
- A finished batch, or a failed task, now raises a desktop notification, so the tab does not have to stay in front.
- Queued downloads can be appended to a running batch, and a single task can be excluded without cancelling the rest.
Vocal separator
- New tool that splits a track into vocals and instrumental using a local model — no audio leaves the device.
- Preview the first 30 seconds, compare both tracks, then export WAV files.
- Uses GPU acceleration when the browser and device support it, and falls back to CPU.
- Up to 500 MB per file and 10 minutes per run; the model is cached after the first download.
Audio noise remover
- New tool that reduces background noise in voice recordings.
- Preview the first 30 seconds, choose how much of the original is blended back in, and download a WAV.
- Processing runs in a worker thread on your device; nothing is uploaded.
Transcription
- Device selection is automatic: GPU when it is available, CPU otherwise, with the choice shown before a long run.
- Long transcription jobs renew their local storage lease, so a slow run is not interrupted partway through.

