w-okada / voice-changer

リアルタイムボイスチェンジャー Realtime Voice Changer
Other
16.12k stars 1.75k forks source link

[ISSUE]: No way to train voice model to generate .JSON file required to use docker version of MMVC #1193

Open Timolian opened 5 months ago

Timolian commented 5 months ago

Voice Changer Version

docker zundamon

Operational System

Windows 11

GPU

none (intel UHD)

Read carefully and check the options

Model Type

MMVC

Issue Description

I installed the MMVC version that doesn't require a CPU because I wanted a real-time conversion with a low computational load. Despite the provided instructional videos being in Japanese, I followed the tutorials and installed the program. The voice recorder apparently needs a .JSON file to operate (which makes sense with the MMVC system). The voice recorder website link, however, was only featured on the Japanese instruction set. Is there a reason for that? The provided video series documents most of the process well, but completely excludes how to generate .JSON files. I could not find any explanation for this. How would you generate these files?

Application Screenshot

No response

Logs on console

No log available; no errors recorded