Microsoft Open-Sources VibeVoice-ASR, a Long-Form Speech Recognition Model That Runs Locally Across 50+ Languages

Microsoft has released VibeVoice-ASR, an open-source speech recognition model built for long-form audio that runs entirely on a local machine, supports more than 50 languages, and removes the need for paid transcription APIs. The model is available under the permissive MIT license through the company's VibeVoice repository on GitHub and as a downloadable model card on Hugging Face.