arctic nacelle
Hey everyone! 👋
I'm trying to run the w-okada/voice-changer completely on cloud GPU
(free tier) since my local PC (i5-6500) can't handle real-time Okada voice changer
voice conversion well.
WHAT I'M TRYING TO DO:
• Run Okada Voice Changer on Kaggle/Colab (free GPU)
• Expose it via ngrok so I can access the WebUI from my browser
• Route the output to VB-Cable for Discord/PUBG
• Use my own .pth RVC voice models
PROBLEMS I'VE HIT:
• Session crashes during heavy pip installs
• Server doesn't always start properly
• ngrok tunnel connects but shows connection refused
• The repo structure is complex with many subfolders
MY SETUP:
• Kaggle with GPU T4 x2 (free tier)
• ngrok for tunneling (free HTTP)
• VB-Cable for virtual audio routing
• Custom .pth voice models
Has anyone successfully done this? Is there a working guide or
notebook that actually runs end-to-end?
Any help would be amazing! Thanks! 🙏
Links:
• Repo: https://github.com/w-okada/voice-changer
• My voice models: RVC .pth files
• Goal: Real-time voice for Discord
next sentinel