Source
Demucs official documentation
Read the primary source.
Reference illustration, not a screenshot of this source.
Demucs
Source-separation model · Voice
Separates audio sources such as vocals and accompaniment.
Use it for
Use before transcription or mixing when music masks speech.
Choose something else for
It does not transcribe or generate narration.
Give it
Mixed audio
Get back
Separated audio stems
How it is operated
Local model or an application such as voice-pro
Availability
Provider capabilities and historical Studio notes are distinguished below. Current account access, credits and runtime readiness were not tested.
How charging works
Compute and model/weight terms; hosted wrappers charge separately.
Where processing happens
Self-hosted if deployed locally, otherwise provider-dependent.
Language support
Exact model version determines support.
Isolate voice from mixed audio
★★★★★ 5/5 for this task
The right specialised layer for vocal separation; compare stem quality on the actual recording.
Decision: Use before transcription or mixing when music masks speech.
Limits and things to check
- Separation can introduce artefacts; it cannot recover every obscured word.
It does not transcribe or generate narration. Separation can introduce artefacts; it cannot recover every obscured word.
Evidence
Official capability documentation; no listening benchmark
Catalogue review: 2026-09-05. The evidence note identifies what was verified. This date does not imply a fresh tool test or verification of every linked source.
