1 Pull request merged by 1 person

• Fixed an issue where onnxruntime would not install on Windows

  #1036 merged Mar 10, 2025

1 Pull request opened by 1 person

• Dev/use vllm

  #1053 opened Mar 7, 2025

26 Issues closed by 5 people

• 是否支持声音文件转声音文件

  #882 closed Mar 12, 2025

• soundfile.LibsndfileError: Error opening 'zero_shot_prompt.wav': System error.

  #883 closed Mar 12, 2025

• English...

  #928 closed Mar 12, 2025

• 【欢迎来到直播间的新朋友】声音克隆这句话,会将【直播间】这个词读取成港普

  #931 closed Mar 12, 2025

• 如何解决Synthesis text segment too short (12) compared to prompt (46), may affect performance这样的性能问题？

  #934 closed Mar 12, 2025

• 自然语言控制不能運作

  #885 closed Mar 11, 2025

• gradio 语音传文字不自动支持，但是huggingface上的demo是支持的，能不能更新下呢？

  #888 closed Mar 11, 2025

• 请问，如何指定GPU，如何指定那个GPU。如果要指定CPU，如何指定。

  #927 closed Mar 11, 2025

• cosyvoice2 inference_zero_shot on Mac

  #894 closed Mar 10, 2025

• How can I train a model with pure speech tokens and text tokens?

  #920 closed Mar 10, 2025

• 使用Cosyvoice2进行情感语音合成

  #921 closed Mar 10, 2025

• 添加英文语气的zero_shot_prompt

  #922 closed Mar 10, 2025

• 加载模型时非常慢，一直在等待，卡在load_hyperpyyaml（）这一步

  #1055 closed Mar 9, 2025

• 关于cosyvoice2.0流式生成，有爆音存在

  #895 closed Mar 9, 2025

• how to train a cosyvoice2 Supervised speech tokenizer with FSQ?

  #896 closed Mar 9, 2025

• 重要！在cosyvoice2 中使用instruct模式的时候，当不传入指令词，效果等同于音色克隆吗？

  #899 closed Mar 8, 2025

• 生成的音频中会混入instruct的指令发音

  #904 closed Mar 8, 2025

• 运行示例代码报错 - Segmentation fault

  #906 closed Mar 8, 2025

• 克隆声音时报错，userwarning:trying to convert audio automatically from float64 to 16-bit int format.

  #1049 closed Mar 7, 2025

• Zero shot 推理中，有概率发生输出开头混入 prompt speech 结尾声音。

  #907 closed Mar 7, 2025

• 日语和韩语生成的音频，有背景音，音乐声

  #908 closed Mar 7, 2025

• 哥哥们，啥时候更新，豆包最新语音都已经卷爆了

  #913 closed Mar 7, 2025

• 关于停顿

  #916 closed Mar 7, 2025

• 求Flow模型在SeedTTS中文测试集上的测试结果

  #1019 closed Mar 7, 2025

• Why is speed adjustment done on mel rather than on wav？

  #1021 closed Mar 6, 2025

• Questions about flow matching decoders

  #905 closed Mar 6, 2025

21 Issues opened by 20 people

• matcha. modles 模块在哪

  #1068 opened Mar 12, 2025

• llm+flow微调

  #1067 opened Mar 12, 2025

• Flow-matching model fine-tuning problem

  #1066 opened Mar 11, 2025

• Finetune CosyVoice-v1-300M with '[accent]' instruct

  #1065 opened Mar 11, 2025

• tn模块在哪里啊？

  #1064 opened Mar 11, 2025

• 翻译

  #1063 opened Mar 11, 2025

• Flow训练报错

  #1062 opened Mar 11, 2025

• 对单个人声音训练总是失败，什么原因

  #1061 opened Mar 11, 2025

• regarding cosyvoice base model training from scratch

  #1060 opened Mar 10, 2025

• 单卡训练报错

  #1059 opened Mar 10, 2025

• text.cc: festival_Text_init open voice lang map failed怎么修改

  #1058 opened Mar 10, 2025

• 在自己生成的音色中instruct情感不明显

  #1057 opened Mar 10, 2025

• CosyVoice-300M-Instruct 可以使用自己复刻的声音吗

  #1054 opened Mar 8, 2025

• Instruction in english

  #1052 opened Mar 7, 2025

• 修复cosyvoice2的flow合成前后爆音

  #1051 opened Mar 7, 2025

• Prompt音频最佳配置是怎样的（会利于TTS效果），有没有有用的经验？

  #1050 opened Mar 7, 2025

• 生成的音频有的音频前部分没声音也不是静音。

  #1048 opened Mar 7, 2025

• 空闲时未自动释放显存 GPU memory is not automatically released in idle state

  #1047 opened Mar 6, 2025

• CosyVoice2 install on windows 10

  #1046 opened Mar 6, 2025

• reason of averaging model

  #1045 opened Mar 6, 2025

• 多线程推理出现错误

  #1044 opened Mar 6, 2025

38 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• llm从头训练，梯度为NAN

  #1032 commented on Mar 6, 2025 • 0 new comments

• 注释存在误解

  #1035 commented on Mar 6, 2025 • 0 new comments

• need to add pyarrow

  #1041 commented on Mar 6, 2025 • 0 new comments

• 使用300M和300M-25Hz权重测试vc接口，合成的音频为静音音频

  #1042 commented on Mar 6, 2025 • 0 new comments

• The embedding shouldn't exist when training LLM for CosyVoice2.0

  #1024 commented on Mar 6, 2025 • 0 new comments

• TTS在4090运行为啥比L20更慢？

  #1022 commented on Mar 6, 2025 • 0 new comments

• 现在都支持哪些方言呢？

  #1018 commented on Mar 6, 2025 • 0 new comments

• Fine-grained的text可不可以生成，而不是手动添加

  #1016 commented on Mar 6, 2025 • 0 new comments

• The Japanese sample audio for “Zero-Shot In-Context Generation” does not match the context.

  #1043 commented on Mar 6, 2025 • 0 new comments

• add another special_tokens

  #1033 commented on Mar 6, 2025 • 0 new comments

• 长文本通过切分增大batch size后并行处理，与串行相比总耗时几乎不变

  #1029 commented on Mar 6, 2025 • 0 new comments

• Regarding to 2.0 inference function.

  #1028 commented on Mar 6, 2025 • 0 new comments

• 如何保存和调用自定义VoiceID？

  #918 commented on Mar 6, 2025 • 0 new comments

• 我使用了server.py文件，并成功部署后，调用接口返回为乱码，该怎么解决？

  #989 commented on Mar 6, 2025 • 0 new comments

• 微调llm

  #1039 commented on Mar 6, 2025 • 0 new comments

• Using phoneme data for CosyVoice

  #1023 commented on Mar 7, 2025 • 0 new comments

• CosyVoice 2.0用4090生成10个字需要4.5秒

  #1003 commented on Mar 7, 2025 • 0 new comments

• Few issues of the dev/lyuxiang.lx branch

  #1017 commented on Mar 7, 2025 • 0 new comments

• 当入参text_frontend=False时，生成的音频会乱说

  #897 commented on Mar 8, 2025 • 0 new comments

• CosyVoice2-0.5B中llm.pt和CosyVoice-BlankEN文件下的模型权重是一致的吗？

  #944 commented on Mar 9, 2025 • 0 new comments

• 谁帮我看看这个？ 十分头疼

  #941 commented on Mar 9, 2025 • 0 new comments

• 调了两个通宵，测了5种语言，不管怎么改都会吞字，弃坑

  #917 commented on Mar 9, 2025 • 0 new comments

• grpc多进程会不会实例化多个模型

  #1025 commented on Mar 10, 2025 • 0 new comments

• cosyvoice = CosyVoice2(CosyVoice2_path, load_jit=True, load_trt=True, fp16=False) load_trt是tersorrt ?

  #1027 commented on Mar 10, 2025 • 0 new comments

• 不好意思我是mac m1 跑出FileNotFoundError的錯誤

  #1031 commented on Mar 10, 2025 • 0 new comments

• How to add other language to CosyVoice2

  #830 commented on Mar 10, 2025 • 0 new comments

• docker创建镜像实在是太大了

  #946 commented on Mar 11, 2025 • 0 new comments

• 如何将CosyVoice2-0.5B模型转成flow.decoder.estimator.fp16.Volta.plan

  #929 commented on Mar 11, 2025 • 0 new comments

• 是否支持训练新的方言，比如中国的闽南语？

  #332 commented on Mar 11, 2025 • 0 new comments

• 请问对 mac mini m4 支持吗，生成的时候 CPU100%，而 GPU 则在旁观望，没有负载。

  #1011 commented on Mar 11, 2025 • 0 new comments

• bitstream多次推理同一段文本后合成结果错乱

  #980 commented on Mar 11, 2025 • 0 new comments

• 关于fastapi/server.py文件下接口参数的问题

  #984 commented on Mar 11, 2025 • 0 new comments

• pip install -r requirements.txt -i https://mirrors.aliyun.com/pypi/simple/ --trusted-host=mirrors.aliyun.com

  #1040 commented on Mar 12, 2025 • 0 new comments

• 看起来大家对流式语音的生成以及长文字tts的速度不满, 大家有优化后的版本么

  #757 commented on Mar 12, 2025 • 0 new comments

• Failed to export onnx model from flow

  #399 commented on Mar 12, 2025 • 0 new comments

• 微调训练报错

  #766 commented on Mar 12, 2025 • 0 new comments

• Add Intel GPU support

  #893 commented on Mar 10, 2025 • 0 new comments

• feat: Support DPO

  #983 commented on Mar 11, 2025 • 0 new comments