BREAKING
Microsoft Open-Sources VibeVoice-ASR
0
min
single-pass audio
0
+
languages
0
B
parameters
0
GB
VRAM
One pass, structured output
1
Speaker diarization
↓
2
Timestamps
↓
3
Transcription
Local model vs cloud APIs
VibeVoice-ASR
MIT
●
Runs offline
●
Zero per-minute cost
●
Global view of audio
Cloud & Whisper
●
Per-minute billing
●
Data leaves device
●
Chunk-based splitting
Research-oriented, use with caution
A new open baseline for speech
AI NEWS BLITZ
Microsoft has released VibeVoice-ASR, an open speech recognition model that runs locally.