Researchers only: Microsoft pulled the long-form speech code after misuse and says it is not for commercial use without further testing.
Kenning's take
What it does
VibeVoice is a family of open voice models from Microsoft Research. The original text-to-speech model could generate up to 90 minutes of speech with up to four speakers; Microsoft later removed that code after misuse, and now offers a smaller real-time streaming speech model plus speech recognition models that transcribe up to an hour of audio with speaker labels. Weights are on Hugging Face, and Microsoft says the models are for research, not commercial use without further testing.
- VibeVoice-Realtime-0.5B streaming text to speech
- Long-form multi-speaker speech up to 90 minutes (original TTS)
- VibeVoice-ASR transcription with speaker labels and timestamps
- Speech recognition in 50+ languages
- Weights on Hugging Face and a Colab demo
Who it suits. Researchers and developers experimenting with open speech synthesis and transcription.
Key facts
- Category
- AI models
- Made by
- Microsoft
- Released
- 25 August 2025
- How to get it
- free open-source code and weights on GitHub and Hugging Face; ASR also in Azure AI Foundry Labs
- Works on
- open-weights
- In the UK
- Yes
- Free plan
- free open-source weights and code
FAQ
When was VibeVoice released?
Microsoft open-sourced VibeVoice-TTS on 25 August 2025, then added a real-time model in December 2025 and speech recognition in January 2026.
Why was VibeVoice-TTS removed?
Microsoft removed the VibeVoice-TTS code on 5 September 2025 after finding it was used in ways inconsistent with its stated research purpose.
Can VibeVoice be used commercially?
Microsoft does not recommend using it in commercial or real-world applications without further testing, and says it is meant for research and development.
Facts from the maker's own pages and the sources listed. Checked 2026-10-11.
Similar tools
ACE-Step 1.5
Open-source music generation model that writes full songs with lyrics and runs on ordinary consumer GPUs.
AlphaGenome Atlas
Google DeepMind database of AI-predicted effects for every possible single-letter change in human DNA.
Amazon Nova 2
Amazon's second family of Nova AI models, covering reasoning, speech conversation and combined text-and-image generation.
Cisco Antares
Cisco's small open-weight models that find where known security vulnerabilities sit in a codebase, running locally.