Microsoft (open weights)
STT TTS 7 models no stock voices based in United States open weights only data from provider API synced 2026-11-18
Microsoft's open-weights speech models on Hugging Face: the VibeVoice ASR and long-form TTS family, Phi-4-multimodal and SpeechT5, mostly MIT-licensed. Microsoft's hosted speech services are listed separately under Azure Speech.
Speech-to-text 4
| Compare | Model | Price |
|---|---|---|
microsoft/Phi-4-multimodal-instructPhi-4-multimodal-instruct |
– | |
microsoft/VibeVoice-ASRVibeVoice-ASR |
– | |
microsoft/VibeVoice-ASR-BitNetVibeVoice-ASR-BitNet |
– | |
microsoft/VibeVoice-ASR-Streaming-7BVibeVoice-ASR-Streaming-7B |
– |
Text-to-speech 3
| Compare | Model | Price |
|---|---|---|
microsoft/speecht5_ttsspeecht5_tts |
– | |
microsoft/VibeVoice-1.5BVibeVoice-1.5B |
– | |
microsoft/VibeVoice-Realtime-0.5BVibeVoice-Realtime-0.5B |
– |