VibeVoice-ASR-HF
Microsoft🇺🇸 USMajor lab·
Audioopen weights
Microsoft's VibeVoice-ASR-HF is a speech-to-text model that transcribes up to 60 minutes of audio in one pass and labels each segment with speaker, start and end time, and content. Its sources mention custom hotwords for domain terms.
Summary of the model card
- License
- mit
- Modalities
- text → text
1 variant
| Standard | HF |
Sources
HuggingFaceDeutsch
VibeVoice-ASR-HF ist ein Speech-to-Text-Modell, das bis zu 60 Minuten Audio in einem Durchgang transkribiert und jedes Segment mit Sprecher, Start- und Endzeit sowie Inhalt ausgibt. Laut Modelcard lassen sich eigene Hotwords fuer Fachbegriffe angeben.
Benchmarks
| Benchmark | Domain | Value | Evidence |
|---|---|---|---|
| Common Voice (ASR)Open ASR Leaderboard | audio-stt | 23.85 % WER | 52 |
| Common Voice (ASR)Open ASR Leaderboard | audio-stt | 20.97 % WER | 52 |
| Common Voice (ASR)Open ASR Leaderboard | audio-stt | 20.6 % WER | 52 |
| Common Voice (ASR)Open ASR Leaderboard | audio-stt | 11.87 % WER | 52 |
| FLEURSOpen ASR Leaderboard | audio-stt | 11.3 % WER | 52 |
| FLEURSOpen ASR Leaderboard | audio-stt | 7.44 % WER | 52 |
| FLEURSOpen ASR Leaderboard | audio-stt | 5.89 % WER | 52 |
| FLEURSOpen ASR Leaderboard | audio-stt | 5.67 % WER | 52 |
| FLEURSOpen ASR Leaderboard | audio-stt | 3.58 % WER | 52 |
Vendor-reported numbers are marked and count with factor 0.7 in rankings. The evidence score says how much a benchmark still tells you today.
Provenance
- Detected
- Aug 25, 2026
- First source
- openasr
- Description
- Summary of the model card
- Editorially reviewed
- —