Yggmicrosoft/VibeVoice
↗ GitHubOpen-Source Frontier Voice AI
34,899
Stars
3,969
Forks
202
Watchers
146
Open Issues
Safety Rating A
No hardcoded secrets, malicious code patterns, suspicious dependencies, or prompt injection attempts were detected. The repository is a well-documented Microsoft Research open-source project under MIT license. The README itself includes a responsible-use disclaimer and notes that the TTS code was removed after discovering misuse cases (deepfakes), demonstrating active responsible-AI governance. No red flags were identified in the repository content provided.
ℹAI-assisted review, not a professional security audit.
AI Analysis
VibeVoice is a family of open-source frontier voice AI models from Microsoft, encompassing Text-to-Speech (TTS), Automatic Speech Recognition (ASR), and real-time streaming TTS capabilities. The ASR model (7B) handles up to 60-minute long-form audio in a single pass with speaker diarization, timestamps, and customized hotword support across 50+ languages. The TTS model (1.5B) generates up to 90 minutes of expressive multi-speaker conversational audio. The streaming TTS model (0.5B) provides real-time synthesis with ~300ms first-audio latency. All models use continuous speech tokenizers at 7.5 Hz and a next-token diffusion framework built on top of a large language model backbone.
Use Cases
- Long-form audio transcription with speaker diarization and timestamps
- Podcast and multi-speaker dialogue synthesis
- Real-time text-to-speech for voice interfaces and input methods
- Multilingual speech recognition and synthesis (50+ languages)
- Fine-tuning speech recognition models on custom domain data
- Voice-powered input methods and accessibility tools
Tags
Project Connections