CosyVoice V3版 - 阿里最新开源语音克隆、文本转语音项目 新增音色融合、字幕同步，本地一键整合包下载

无言***

Arwar 发表于 2024-9-4 14:43
调速出现以上错误，不调就正常

这是生成音频，还是克隆？

yh***

好东西

无言***

Arwar 发表于 2024-9-4 14:42
Failed to change speed of audio:
[WinError 2] 系统找不到指定的文件。
Traceback (most recent call l ...

To create a public link, set `share=True` in `launch()`
从这段开始，下面的复制，贴出来

海南***

我要下载最新版，我又来顶帖了。

Ar***

预训音色和3秒极速复刻，只要调了语速就会报错，不调就正常


To create a public link, set `share=True` in `launch()`.
2024-09-05 09:58:42,843 INFO get sft inference request
Building prefix dict from the default dictionary ...
2024-09-05 09:58:43,173 DEBUG Building prefix dict from the default dictionary ...
Loading model from cache C:\Users\MSI-NB\AppData\Local\Temp\jieba.cache
2024-09-05 09:58:43,174 DEBUG Loading model from cache C:\Users\MSI-NB\AppData\Local\Temp\jieba.cache
Loading model cost 0.409 seconds.
2024-09-05 09:58:43,582 DEBUG Loading model cost 0.409 seconds.
Prefix dict has been built successfully.
2024-09-05 09:58:43,582 DEBUG Prefix dict has been built successfully.
['轻轻的我走了，正如我轻轻地来。我挥一挥衣袖，不带走一片云彩。、。']
轻轻的我走了，正如我轻轻地来。我挥一挥衣袖，不带走一片云彩。、。
ZH1
读取pt
融合音色: 团长_悲伤
C:\CosyVoice-V3\python\Lib\site-packages\diffusers\models\attention_processor.py:1279: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.)
  hidden_states = F.scaled_dot_product_attention(
tensor([[ 7.3901e-06,  4.5658e-06,  1.3181e-05,  ..., -1.4055e-03,
         -1.4441e-03, -1.4148e-03]])
[ 7.3901201e-06  4.5658371e-06  1.3180562e-05 ... -1.4055450e-03
-1.4440864e-03 -1.4147667e-03]
[tensor([[ 7.3901e-06,  4.5658e-06,  1.3181e-05,  ..., -1.4055e-03,
         -1.4441e-03, -1.4148e-03]])]
推理方法 inference_sft 运行时长: 6.9186 秒
Failed to change speed of audio:
[WinError 2] 系统找不到指定的文件。
Traceback (most recent call last):
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\queueing.py", line 536, in process_events
    response = await route_utils.call_process_api(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\route_utils.py", line 276, in call_process_api
    output = await app.get_blocks().process_api(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\blocks.py", line 1897, in process_api
    result = await self.call_function(
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\blocks.py", line 1483, in call_function
    prediction = await anyio.to_thread.run_sync(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\anyio\to_thread.py", line 56, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\anyio\_backends\_asyncio.py", line 2177, in run_sync_in_worker_thread
    return await future
           ^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\anyio\_backends\_asyncio.py", line 859, in run
    result = context.run(func, *args)
             ^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\utils.py", line 816, in wrapper
    response = f(*args, **kwargs)
               ^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\webui.py", line 240, in generate_audio
    return (target_sr, audio_data)
                       ^^^^^^^^^^
UnboundLocalError: cannot access local variable 'audio_data' where it is not associated with a value

无言***

Arwar 发表于 2024-9-5 10:04
预训音色和3秒极速复刻，只要调了语速就会报错，不调就正常

用系统自带的女声音色合成试试
就是启动后，只调节加速，然后生成，其他默认

Ar***

To create a public link, set `share=True` in `launch()`.
2024-09-05 10:21:06,776 INFO get sft inference request
Building prefix dict from the default dictionary ...
2024-09-05 10:21:07,088 DEBUG Building prefix dict from the default dictionary ...
Loading model from cache C:\Users\MSI-NB\AppData\Local\Temp\jieba.cache
2024-09-05 10:21:07,089 DEBUG Loading model from cache C:\Users\MSI-NB\AppData\Local\Temp\jieba.cache
Loading model cost 0.425 seconds.
2024-09-05 10:21:07,514 DEBUG Loading model cost 0.425 seconds.
Prefix dict has been built successfully.
2024-09-05 10:21:07,517 DEBUG Prefix dict has been built successfully.
['轻轻的我走了，正如我轻轻地来。我挥一挥衣袖，不带走一片云彩。、。']
轻轻的我走了，正如我轻轻地来。我挥一挥衣袖，不带走一片云彩。、。
C:\CosyVoice-V3\python\Lib\site-packages\diffusers\models\attention_processor.py:1279: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.)
  hidden_states = F.scaled_dot_product_attention(
tensor([[ 0.0345,  0.0257,  0.0137,  ..., -0.0108,  0.0216,  0.0388]])
[ 0.03447248  0.02565259  0.01370092 ... -0.01083746  0.02156586
  0.03875655]
[tensor([[ 0.0345,  0.0257,  0.0137,  ..., -0.0108,  0.0216,  0.0388]])]
推理方法 inference_sft 运行时长: 6.7179 秒
Failed to change speed of audio:
[WinError 2] 系统找不到指定的文件。
Traceback (most recent call last):
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\queueing.py", line 536, in process_events
    response = await route_utils.call_process_api(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\route_utils.py", line 276, in call_process_api
    output = await app.get_blocks().process_api(
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\blocks.py", line 1897, in process_api
    result = await self.call_function(
             ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\blocks.py", line 1483, in call_function
    prediction = await anyio.to_thread.run_sync(
                 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\anyio\to_thread.py", line 56, in run_sync
    return await get_async_backend().run_sync_in_worker_thread(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\anyio\_backends\_asyncio.py", line 2177, in run_sync_in_worker_thread
    return await future
           ^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\anyio\_backends\_asyncio.py", line 859, in run
    result = context.run(func, *args)
             ^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\python\Lib\site-packages\gradio\utils.py", line 816, in wrapper
    response = f(*args, **kwargs)
               ^^^^^^^^^^^^^^^^^^
  File "C:\CosyVoice-V3\webui.py", line 240, in generate_audio
    return (target_sr, audio_data)
                       ^^^^^^^^^^
UnboundLocalError: cannot access local variable 'audio_data' where it is not associated with a value

启动后，女生，加速1.2，生成，还是出错

无言***

Arwar 发表于 2024-9-5 10:22
To create a public link, set `share=True` in `launch()`.
2024-09-05 10:21:06,776 INFO get sft infer ...

回头我看下是不是打包漏掉了文件

po***

谢谢分享~~~~~~~

wangh***

看看作者辛苦