[GH-ISSUE #687] Voice changing issue (CUDA out of memory) #414

New issue

Closed

opened 2026-10-20 09:59:07 +00:00 by bsakaguchi · 4 comments

bsakaguchi commented

2026-10-20 09:59:07 +00:00

Owner

Originally created by @EnderRL on GitHub (Jul 4, 2023).
Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/687

Hi I am new to RVC, and I have little knowledge about it.

I was trying out how powerful the application, and it was really cool. I was able to separate the voice and instrumental.
But when it comes to changing voice, I getting these error:

I assume that because my gpu ran out of memory:

I was using a random trained voice (the .pth file) from this repo
And I sure it wasn’t the problem (I haven’t try the other voice yet), and it was because of my gpu limitations. I also haven’t try train a voice yet but I assume that it will also give out of memory error.

So my question is:

Can I solve the out of memory issue? Was it related to the “config.py”? Because I did read the faq and I think I need to change some settings in the file.
If not, can I use cpu processing for the voice changer? I am fine with the slow processing speed as long as it came with results!

Pc specs:
Cpu: Core i3 gen 12th (4 core 8 threads)
Ram: 16GB
Gpu: GTX 950 (2GB vram)

I also forgot to mention that I used the installation script from this video (or the powershell iex script: iex (irm rvc.tc.ht)) so that I saved my time installing manually.

Thanks!

哈咯哈咯，我的华文会很差，只是希望可以让你们明白点。
我刚接触这个变音器，所以我不是很会用。

我只是来看看这个东西有多好玩，就觉得这个分唱歌的和伴奏的很成功。
但是到变音的时候，就开始有问题。我是觉得因为out of memory的问题，所以就求方法搞这个问题。

所以我想问的是请问那个out of memory的问题可以解决吗？是不是关系到“config.py”？
如果不能的话我可以用CPU操作吗？

我电脑的配件:
Cpu: intel core i3 gen 12th
Ram: 16GB
Gpu: GTX 950 (2GB vram)

謝謝!

Originally created by @EnderRL on GitHub (Jul 4, 2023). Original GitHub issue: https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/issues/687 Hi I am new to RVC, and I have little knowledge about it. I was trying out how powerful the application, and it was really cool. I was able to separate the voice and instrumental. But when it comes to **changing voice**, I getting these error: ![image](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/assets/121793770/8b3e3038-5168-4efb-941f-3a232e75c116) I assume that because my gpu **ran out of memory**: ![image](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/assets/121793770/29036aff-552b-419a-aff6-a6d1cdda0426) I was using a random trained voice (the .pth file) from this [repo](https://huggingface.co/analogspiderweb/analogspiderweb_RVC/blob/main/villager.zip) And I sure it wasn’t the problem (I haven’t try the other voice yet), and it was because of my gpu limitations. I also haven’t try train a voice yet but I assume that it will also give out of memory error. So my question is: 1. Can I solve the out of memory issue? Was it related to the “[config.py](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/config.py)”? Because I did read the [faq](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/wiki/FAQ-(Frequently-Asked-Questions)#q8cuda-errorcuda-out-of-memory) and I think I need to change some settings in the file. 2. If not, can I use cpu processing for the voice changer? I am fine with the slow processing speed as long as it came with results! Pc specs: Cpu: Core i3 gen 12th (4 core 8 threads) Ram: 16GB Gpu: GTX 950 (2GB vram) I also forgot to mention that I used the installation script from this [video](https://youtu.be/Y8IxVVQBEpc) (or the powershell iex script: `iex (irm rvc.tc.ht)`) so that I saved my time installing manually. Thanks! 哈咯哈咯，我的华文会很差，只是希望可以让你们明白点。我刚接触这个变音器，所以我不是很会用。我只是来看看这个东西有多好玩，就觉得这个分唱歌的和伴奏的很成功。但是到变音的时候，就开始有问题。我是觉得因为out of memory的问题，所以就求方法搞这个问题。 1. 所以我想问的是请问那个out of memory的问题可以解决吗？是不是关系到“config.py”？ 2. 如果不能的话我可以用CPU操作吗？我电脑的配件: Cpu: intel core i3 gen 12th Ram: 16GB Gpu: GTX 950 (2GB vram) 謝謝!

bsakaguchi

2026-10-20 09:59:07 +00:00

closed this issue
added the
following up
label

bsakaguchi commented

2026-10-20 09:59:07 +00:00

Author

Owner

@EnderRL commented on GitHub (Jul 4, 2023):

Update: nevermind, in the "config.py", just lower the x_pad, x_query, x_center, x_max

so it won't show cuda timeout, and the problem is not even related to gpu vram, it was fine.
Let me know if i can actually improve this, otherwise problem solved! Thanks!

@EnderRL commented on GitHub (Jul 4, 2023): Update: nevermind, in the "[config.py](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/config.py)", just lower the `x_pad, x_query, x_center, x_max` ![image](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/assets/121793770/e4857a8d-43d0-45f6-b0bd-1956cd0ee4a9) so it won't show cuda timeout, and the problem is not even related to gpu vram, it was fine. Let me know if i can actually improve this, otherwise problem solved! Thanks!

bsakaguchi commented

2026-10-20 09:59:07 +00:00

Author

Owner

@ms903x1 commented on GitHub (Jul 4, 2023):

gtx950 is not within the supported range. Please use a Niadiv graphics card of gtx10xx series or higher

@ms903x1 commented on GitHub (Jul 4, 2023): gtx950 is not within the supported range. Please use a Niadiv graphics card of gtx10xx series or higher

bsakaguchi commented

2026-10-20 09:59:07 +00:00

Author

Owner

@RVC-Boss commented on GitHub (Jul 5, 2023):

2G GPU memory is not enough for inference.
If you want to use CPU to inference, you can modify this line
self.device = "cuda:xxx"
to
self.device = "cpu"

@RVC-Boss commented on GitHub (Jul 5, 2023): 2G GPU memory is not enough for inference. If you want to use CPU to inference, you can modify this line self.device = "cuda:xxx" to self.device = "cpu"

bsakaguchi commented

2026-10-20 09:59:08 +00:00

Author

Owner

@EnderRL commented on GitHub (Jul 6, 2023):

Thanks! It works very stable for now.

@EnderRL commented on GitHub (Jul 6, 2023): > Thanks! It works very stable for now.