Tongyi Speech (CosyVoice)
Alibaba · China · Speech & audio
China Open sourceFree tier
Showing version: Fun-CosyVoice 3.0（0.5B） · 2025-12-01 · Version lineage ↓
01
Overview
Alibaba Tongyi speech lab's (FunAudioLLM) open-source speech generation family, centred on the CosyVoice TTS series: multilingual synthesis, voice cloning from ~3 seconds of reference audio and streaming generation, released under Apache-2.0; the sibling SenseVoice model covers speech understanding/ASR.
Release date2025-12-01
02 / Provenance
Field checks and source notes
When sources disagree on a field, every reading is listed here with its own citation. We do not take a midpoint, invent a missing figure, or decide which side is right — follow the citation and judge for your case.
These field-level notes are recorded verbatim in Chinese in the source data. Open the Chinese version
03 / Plain layer
Plain-language benchmarks
Naturalness Not disclosed
Accuracy (WER) Not disclosed
Speed / latency Not disclosed
Languages supported Not disclosed
Price / free Not disclosed开源权重免费自部署
04
Practical facts
| Context length | Not disclosed |
|---|---|
| Max output | Not disclosed |
| Parameters | Not disclosed |
| Input | Not disclosed |
| Output price | Not disclosed |
| CNY | Open weights can be downloaded free for self-hosting under Apache-2.0; cloud API pricing on Alibaba Cloud Bailian and similar platforms follows the official pages (CNY list prices, per the vendor's official pricing page). |
| Free tier | Open weights, free to self-host. |
| Output speed | Not disclosed |
| Release date | |
| License | Not disclosed |
Show technical data
Technical benchmark table
No verified benchmark data (not tested / not disclosed / not applicable).
| Version | Release date | Status | Openness | Parameters | What changed | Detail | |
|---|---|---|---|---|---|---|---|
| Fun-CosyVoice 3.0（0.5B）Showing version | Active | Open source | Not disclosed | Third generation, open-sourced December 2025; improves content consistency, speaker similarity and prosody over 2.0, with 9 languages and 18+ Chinese dialects per the HF model page and project materials. | Detail | ||
| CosyVoice 2.0（CosyVoice2-0.5B） | Superseded | Open source | Not disclosed | Released December 2024 with a streaming-first redesign and ~150 ms first-packet latency per project materials. | Detail | ||
| CosyVoice 1.0 | Superseded | Open source | Not disclosed | First released July 2024 alongside SenseVoice under the FunAudioLLM project. | Detail |
Last updated: · Sources on detail page
06 / Compare