ASR comparison
VibeVoice-ASR next to FunASR and Whisper — open-source speech-to-text, side by side. Updated for the streaming and CPU builds.
VibeVoice-ASR 7B long-form | FunASR open Alibaba | Whisper open OpenAI · open | gpt-4o-transcribe API OpenAI · API | |
|---|---|---|---|---|
| Open license | MIT | MIT | MIT | Proprietary |
| Long-form single pass | 60 min | check | chunked | chunked |
| Speaker diarization | check | check | — | — |
| Timestamps | check | check | check | limited |
| Streaming / real-time | since Sep 2026 | check | — | check |
| Runs on CPU (no GPU) | BitNet build | check | whisper.cpp | — |
| Languages | 50+ (10 streaming) | 50+ | ~99 | Multi |
| Speed | 60-min pass | up to 170x | baseline | high accuracy |
| Self-host / local | check | check | check | — |