Realtime voice

Low-latency streaming voice — VibeVoice-Realtime next to OpenAI's latest realtime model and CosyVoice 2.

VibeVoice-Realtime
0.5B
open · self-host
GPT-Realtime-2
API
OpenAI · 2026
CosyVoice 2
Apache-2.0
Alibaba
Typestreaming TTSspeech-to-speechstreaming TTS
First-audio latency~300 mslow~150 ms
Streaming inputcheckcheckcheck
Reasoning / dialoguecheck
Live translationcheck
Open / self-hostcheckcheck
LicenseMITProprietaryApache-2.0
Context window8K128K