VibeVoice-ASR-Streaming

Microsoft🇺🇸 USMajor lab· Sep 2, 2026

Audioopen weights

Streaming ASR model that transcribes speech as it arrives and attributes each utterance to a speaker (who said what). Users can supply custom hotwords such as names and technical terms, and the card lists ten languages.

Summary of the model card

License: mit
Modalities: audio → text

2 variants

1.5B	1.5B			HF
7B	7B			HF

Sources

HuggingFace Paper

Deutsch

Streaming-ASR-Modell, das Sprache während der Eingabe transkribiert und dabei Sprecher zuordnet (wer sagte was). Eigene Hotwords wie Namen und Fachbegriffe verbessern die Erkennung, laut Modelcard werden zehn Sprachen unterstützt.

Provenance

Detected: Sep 27, 2026
First source: huggingface
Description: Summary of the model card
Editorially reviewed: —