search

SociallyIneptWeeb / AICoverGen

A WebUI to create song covers with any RVC v2 trained AI voice from YouTube videos or audio files.
MIT License
1.13k stars 267 forks source link

How to infer song using only CPU? #127

Open anywindo opened 7 months ago

anywindo commented 7 months ago

I don't have NVIDIA GPU so, i tweak the os environment to set visible cuda device to -1, it raised an error "Torch not compiled with cuda enabled". How can i do it using only CPU?

Dimkqo commented 6 months ago

I solved that problem in part. There some code problems in module mdx.py. So, i edited these lines and now it can separate vocals and instrumental:

mdx.py: ... 169 mix_waves = torch.tensor(mix_waves, dtype=torch.float32)#.to(self.device) ... 194 processed_wav = self.model.istft(processed_spec)#.to(self.device)) ... 245 #device_properties = torch.cuda.get_device_properties(device) if device.type == 'cuda' else torch.cpu.get_device_properties(device) 246 #vram_gb = device_properties.total_memory / 1024**3 247 m_threads = 1 #if vram_gb < 8 else 2

Perhaps, even can separate lead and back vocals and dereverb main vocal. But i can't change the voice line in this moment. I hope i can fix this soon.

Edited: I use console version without web ui.