Issues

• Verify evals on Papers with Code

  QwenAudio/CosyVoice#1956 · Filter by author NielsRogge opened on Oct 12, 2026
  2

• Invitation to evaluate CosyVoice with RVCBench datasets and pip tools

  QwenAudio/CosyVoice#1955 · Filter by author Nanboy-Ronan opened on Oct 10, 2026
  1

• CosyVoice 2/3 output collapses into unintelligible speech on transformers >= 4.53: forward_one_step passes a mask frozen at prompt length

  QwenAudio/CosyVoice#1953 · Filter by author xuyukuan opened on Oct 1, 2026
  1

• Streaming: LM and flow decoder on separate processes are ~1.4x faster

  QwenAudio/CosyVoice#1951 · Filter by author shrut2702 opened on Sep 25, 2026
  2

• text_normalize in cosyvoice/cli/frontend.py works only for Chinese/English

  QwenAudio/CosyVoice#1949 · Filter by author ossk opened on Sep 21, 2026
  1

• https://funaudiollm.github.io/cosyvoice3页面无法打开404

  stale
  QwenAudio/CosyVoice#1947 · Filter by author fdfbest opened on Sep 15, 2026
  2

• WER of token2text

  QwenAudio/CosyVoice#1946 · Filter by author anshkumar opened on Sep 14, 2026
  3

• License clarification: using Fun-CosyVoice3 outputs to train and sell an independent real-time VC mode

  stale
  QwenAudio/CosyVoice#1945 · Filter by author Urusibara-Kamatari opened on Sep 11, 2026
  2

• CosyVoice2: is multi-utterance batch (B>1) inference supported or planned?

  stale
  QwenAudio/CosyVoice#1944 · Filter by author James-Bond123 opened on Sep 9, 2026
  2

• CosyVoice3推理加速优化方案：ONNX图优化、TensorRT推理调整、分桶策略与CUDA Graph

  QwenAudio/CosyVoice#1892 · Filter by author BeckYang26 opened on May 19, 2026
  56

• CosyVoice 2.0 exhibits audio stuttering during zero-shot streaming inference

  QwenAudio/CosyVoice#1740 · Filter by author HUSTHY opened on Dec 26, 2025
  19

• 方言微调cosyvoice2求助

  stale
  QwenAudio/CosyVoice#1653 · Filter by author lunar333 opened on Nov 21, 2025
  18

• 生成的音频是胡言乱语

  QwenAudio/CosyVoice#1481 · Filter by author WangGewu opened on Jul 24, 2025
  22

• 求助，下载git模型的Cosyvoice2-0.5B，在comfyui运行后提示缺失文件

  QwenAudio/CosyVoice#1276 · Filter by author Exit-er opened on May 10, 2025
  11

• Need amp_training config and pipeline on flow model

  stale
  QwenAudio/CosyVoice#1205 · Filter by author JohnHerry opened on Apr 21, 2025
  2

• token2wav开销很大，有什么方向可以优化该处性能

  QwenAudio/CosyVoice#1202 · Filter by author wenyangchou opened on Apr 19, 2025
  2

• 是否支持双向流式输入和输出？

  stale
  QwenAudio/CosyVoice#1201 · Filter by author mmxuan18 opened on Apr 18, 2025
  7

• flow.cache.pt vs flow.pt: Which to use for adaptation in CosyVoice2

  stale
  QwenAudio/CosyVoice#1200 · Filter by author jeongminju0815 opened on Apr 18, 2025
  2

• flow2训练大量warning get chunk_masks all false

  stale
  QwenAudio/CosyVoice#1199 · Filter by author boji123 opened on Apr 18, 2025
  3

• 方言的训练需要怎么做，并可以在CosyVoice中使用？

  stale
  QwenAudio/CosyVoice#1194 · Filter by author MagicalWhirlwind opened on Apr 17, 2025
  2

• ZeroDivisionError: 0.0 cannot be raised to a negative power

  stale
  QwenAudio/CosyVoice#1183 · Filter by author hjj-lmx opened on Apr 16, 2025
  10

• flow match cache版本推理出现爆音，同时prompt文本过长，推理时稳定出现合成音频前面带有prompt中的音频

  stale
  QwenAudio/CosyVoice#1178 · Filter by author donstang opened on Apr 15, 2025
  8

• webui.py的demo演示中，如何添加预训练的音色文件？

  stale
  QwenAudio/CosyVoice#1177 · Filter by author Love-Yuki opened on Apr 15, 2025
  6

• cosyvoice flow2 转 fp16 trt时的异常

  QwenAudio/CosyVoice#1173 · Filter by author JohnHerry opened on Apr 14, 2025
  10

• 使用指导生成，instruct_text 设置为 用中文(普通话)说这句话，生成的音频感觉像是有老外的语气再说这句话，使用三秒复刻就是正常的

  stale
  QwenAudio/CosyVoice#1172 · Filter by author lsh-boop opened on Apr 14, 2025
  2