8 Issues closed by 5 people

• 使用add_zero_shot_spk后，再inference_zero_shot，生成的声音是错乱的

  #1434 closed Jul 7, 2025

• 有试过把Hifigan换成Bigvgan，效果会更好的朋友吗？

  #1306 closed Jul 4, 2025

• vllm测试自己sft的模型缺少参数

  #1415 closed Jul 3, 2025

• campplus.onnx对应的pytorch模型

  #1421 closed Jul 2, 2025

• [Question] Add new emotion tags

  #1275 closed Jul 2, 2025

• Adding languages to multi-lingual and zero shot voice cloning / 為多語言和零樣本語音克隆添加語言

  #1278 closed Jul 1, 2025

• flow 非cache版本的onnx生成

  #1283 closed Jul 1, 2025

• Training details and hallucination handling

  #1390 closed Jul 1, 2025

17 Issues opened by 16 people

• 推理的时候已经保证batch为1了，还要这个mask干嘛啊？

  #1437 opened Jul 7, 2025

• 生成的声音时不时会出现截断。有这个警告Using None for EOS token id because tokenizer is not initialized

  #1436 opened Jul 7, 2025

• 为什么CosyVoice-300M-SFT的加载速度这么慢

  #1435 opened Jul 7, 2025

• no valid model_type

  #1433 opened Jul 4, 2025

• dpo for cosyvoice2-instruct

  #1432 opened Jul 4, 2025

• “-”被读作“减”

  #1431 opened Jul 4, 2025

• 910B

  #1430 opened Jul 3, 2025

• 使用python3.11安装环境，推理时会出现乱读现象

  #1429 opened Jul 3, 2025

• 如何获取训练数据有哪些“方言”，“情感”，“速度”，“角色”相关prefix单词？

  #1428 opened Jul 2, 2025

• Is there any ttsfrd-0.4.2-cp39-cp39-linux_x86_64.whl?

  #1427 opened Jul 2, 2025

• Cosyvoice2使用fastapi对外输出音频,和本地脚本执行转换结果不一样

  #1426 opened Jul 2, 2025

• 上句尾和下句头乱衔接

  #1425 opened Jul 2, 2025

• Cosyvoice2 sft 支持添加自然语言指令进行微调吗

  #1423 opened Jul 2, 2025

• CosyVoice2 CosyVoice only inference_sft success other inference as inference_zero_shot failed

  #1422 opened Jul 2, 2025

• cosyvoice2-0.5B 乱读和重复读问题

  #1420 opened Jul 1, 2025

• 模型应该微调llm还是flow

  #1419 opened Jul 1, 2025

• 安装问题

  #1418 opened Jul 1, 2025

21 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.

• 提高flow模型训练速度和GPU浮点利用率的方法收集

  #1321 commented on Jul 1, 2025 • 0 new comments

• vllm+fp16 推理，inference_vc 有几率生成空的音频或低音量音频。

  #1414 commented on Jul 1, 2025 • 0 new comments

• windwows端推理必报错

  #1341 commented on Jul 2, 2025 • 0 new comments

• 疑似内存泄露问题

  #491 commented on Jul 2, 2025 • 0 new comments

• cosyvoice3

  #1323 commented on Jul 2, 2025 • 0 new comments

• Cosyvoice2 SFT 验证集loss只升不降

  #1397 commented on Jul 2, 2025 • 0 new comments

• 如何避免instruct_text被读出来？

  #1263 commented on Jul 2, 2025 • 0 new comments

• cuda 测试生成很慢,正常吗? 近10秒TTS

  #1339 commented on Jul 3, 2025 • 0 new comments

• CosyVoice2-0.5B流式播放总算正常

  #1396 commented on Jul 3, 2025 • 0 new comments

• docker build failed：timeout

  #1344 commented on Jul 4, 2025 • 0 new comments

• 没有可用的预训练音色！

  #1343 commented on Jul 4, 2025 • 0 new comments

• CosyVoice2的方言口音微調

  #1105 commented on Jul 4, 2025 • 0 new comments

• CosyVoice 2.0 克隆模式下自然语言描述语气功能不工作

  #1308 commented on Jul 4, 2025 • 0 new comments

• 使用text_generator生成器报错，'Qwen2Model' object has no attribute 'layers'

  #1348 commented on Jul 5, 2025 • 0 new comments

• Finetuning steps

  #1336 commented on Jul 5, 2025 • 0 new comments

• vllm加速后会出现吞字漏字的现象怎么解决

  #1353 commented on Jul 6, 2025 • 0 new comments

• Other languages

  #202 commented on Jul 6, 2025 • 0 new comments

• 利用vllm加速llm模型但是生成的是无意义的噪音

  #1316 commented on Jul 7, 2025 • 0 new comments

• `KeyError: 'prompt_text_len'` (之前遇到的 `KeyError: 'prompt_text'`)

  #1358 commented on Jul 7, 2025 • 0 new comments

• 使用指定音色id时报错llm_prompt_speech_token

  #1357 commented on Jul 7, 2025 • 0 new comments

• The latest code is incompatible with the model file version on Hugging Face

  #1407 commented on Jul 7, 2025 • 0 new comments