All tasks / Models
VibeVoice-1.5B
Microsoft's MIT-licensed open-weights text-to-speech model for long-form, multi-speaker podcast-style audio (English, Chinese).
- Microsoft
- Open weights
- Released 26 Aug 2025
Where it ranks
- CText to speechModels · Local · 957 on Artificial Analysis Speech Arena: Text to Speech (provider voices)
Facts
- Vendor
- Microsoft
- Family
- VibeVoice
- Released
- 26 Aug 2025
- Weights
- Open: download and run it yourself
- License
- MIT
- Parameters
- 2.7B
- GPU memory to run locally
- ~8 GB (Derived: 2.7B parameters (Hugging Face safetensors count) x 2 bytes = ~5.4 GB of bf16 weights, plus overhead.)
- Access
- Free (open weights)
- Last verified
- 18 Oct 2026
Benchmark results
| Benchmark | Result | Entry on the leaderboard | As of | Source |
|---|---|---|---|---|
| Artificial Analysis Speech Arena: Text to Speech (provider voices) | 957 | VibeVoice 1.5BProvider Voice Arena, all categories and accents; rank 81 of 96, 2127 appearances, 95% CI ±14; raw value from the page's embedded data (shown rounded); no last-updated date shown, so as_of = retrieval date. | 18 Oct 2026 | artificialanalysis.ai |
Where you can use it
- CTTS WebUILocal servers & APIs · Local