Unable to get Openvoice 2.0 working on windows ( Could not load library cudnn_ops_infer64_8.dll. Error code 126 )

[krylor]

I am working on windows 11.
Trying to get Openvoice 2.0 working and facing this error consistently.
OpenVoice version: v2
Could not load library cudnn_ops_infer64_8.dll. Error code 126
Please make sure cudnn_ops_infer64_8.dll is in your library path!

What version of Cuda and Cuddn do I need to install, I am getting a lot of conflicting errors on this.

[Surya-Git-hub]

hey @krylor i am having same issue, did you find any solution?

[Zephiel13]

I have the same problem. cudnn_ops_infer64_8.dll is in cudnn V8 , it seems that openvoice uses cudnn v8 and cuda 11.x

[krylor]

I just used Google collab finally.

…

On Tue, Aug 20, 2024 at 2:19 AM Zephiel13 ***@***.***> wrote: I have the same problem. cudnn_ops_infer64_8.dll is in cudnn V8 , it seems that openvoice uses cudnn v8 and cuda 11.x — Reply to this email directly, view it on GitHub <#284 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A52HRA5YVXCH3M2SJ564F4DZSJK6JAVCNFSM6AAAAABLGSL6MSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMZYHA2TCOI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[MeachamusPrime]

Download CUDNN v8.9 from NVidia's developer website. Install. Reboot. It should then get past that.

[krylor]

Thanks, I'll check it out.

…

On Sun, Oct 13, 2024 at 8:46 AM Christopher Meacham < ***@***.***> wrote: Download CUDNN v8.9 from NVidia's developer website. Install. Reboot. It should then get past that. — Reply to this email directly, view it on GitHub <#284 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A52HRA3VJKT4E7TZ7V7TBFLZ3HQX3AVCNFSM6AAAAABLGSL6MSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAOJSGU4TGMQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[mytait]

i got this solved for me:
under the hood there is the option to use 2 libaries for audio separation: vad and whisper

i think the undelying library whisper is selected by default but i think broken..

-it does not work in CPU mode
-and even in CUDA its broken.

easy fix: use vad.

For it just

change this line:

target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, vad=False)

and set the vad to true:

target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, vad=True)

then it will use vad and should work.

basically anywhere that the function "get_se" is calles you can add a ", vad=True" parameter to fix it.

Notebook1 and 2 use vad=True and work with no problems but notebook3 has vad=False and is broken.
Just change it to True

hope this helps!

[krylor]

Thanks for the tip!

…

On Sun, Oct 13, 2024 at 8:30 PM Tait ***@***.***> wrote: i got this solved for me: under the hood there is the option to use 2 libaries for audio separation: vad and whisper i think the undelying library whisper is selected by default but i think broken.. -it does not work in CPU mode -and even in CUDA its broken. easy fix: use vad. For it just change this line: target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, vad=False) and set the vad to true: target_se, audio_name = se_extractor.get_se(reference_speaker, tone_color_converter, vad=True) then it will use vad and should work. basically anywhere that the function "get_se" is calles you can add a ", vad=True" parameter to fix it. hope this helps! — Reply to this email directly, view it on GitHub <#284 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A52HRA4JFHMMEAH4CZB52QLZ3KDJHAVCNFSM6AAAAABLGSL6MSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAOJSHA4TGMQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>