search

w-okada / voice-changer

リアルタイムボイスチェンジャー Realtime Voice Changer
Other
15.54k stars 1.68k forks source link

feat: copy and refacotor the code from DDSP_SVC #1017

Open shdancer opened 8 months ago

shdancer commented 8 months ago

I see the feature request from issue #762. Then, I referenced the newest code from DDSP-SVC 4.0 and refactored the DDSP-SVC code. Here are several changes and the reasons:

According to the recommendation of DDSP-SVC, there's no need to pass the DDSP model; only the Diff model is required. So, I changed the frontend code to allow the model to be empty. Referring to gui_diff.py in DDSP-SVC's repository, I modified the settings of the DDSP-SVC voice changer. Known bugs: When the chunk size is too small, it seems that DDSP-SVC itself will encounter an error related to f0_extraction.

This is my first time doing a PR. If @w-okada can help me merge, I would be really apppreciated for that.

github-actions[bot] commented 8 months ago

CLA Assistant Lite bot All contributors have signed the CLA ✍️ ✅

shdancer commented 8 months ago

I have read the CLA Document and I hereby sign the CLA

w-okada commented 8 months ago

we cannot use praat-parselmouth because of license issue.

shdancer commented 8 months ago

So if I remove the praat-parselmouth, will it be safe? In fact, the project is not using parselmouth as a f0_extractor besiedes.

Already remove praat-parselmout @w-okada

yxlllc commented 7 months ago

So if I remove the praat-parselmouth, will it be safe? In fact, the project is not using parselmouth as a f0_extractor besiedes.

Already remove praat-parselmout @w-okada

ddsp-svc is using rmvpe as default f0 extractor now.