Skip to content

Timeline Studio v0.3.0 — Browser AI Vocal Separation

Choose a tag to compare

View all tags
@MartinDelophy MartinDelophy released this 13 Jul 08:02
v0.3.0
e00b2e7

What's new

  • Added fully browser-side AI vocal and instrumental separation.
  • Bundled the UVR model and 16 weight shards for reliable local loading and service-worker caching.
  • Added TensorFlow.js 4.22 WebGPU acceleration with automatic WebGL fallback.
  • Added end-to-end validated STFT, model inference, inverse STFT, and stereo WAV output.
  • Added Chinese and English copy for separation progress, errors, output names, and notifications.
  • Kept hidden timeline clips visible with a disabled treatment instead of removing them.
  • Added independent visibility state for multiple caption and voiceover lanes.
  • Added the modern mobile web app capability metadata.

Validation

  • Production build passes.
  • Timeline tests: 9/9 passing.
  • Browser E2E vocal separation produced valid vocal and instrumental WAV outputs.

Notes

The vocal-separation model is approximately 66 MB and is cached after its first load.

Assets 2
Loading