NaruseMioShirakana / DragonianVoice

多个SVC/TTS的C++推理库
GNU Affero General Public License v3.0
985 stars 119 forks source link

lnvalid Feed lnput Name:xx #25

Closed misss520 closed 1 year ago

misss520 commented 1 year ago

I want to know what causes this error 捕获

NaruseMioShirakana commented 1 year ago

This problem may be caused by not specifying V3/V4 in the model configuration file. If you are using SoVits3.0, add "SoVits3": true to the model configuration file. If you are using SoVits4.0, add "SoVits4": true to the model configuration file, If you are using SoVits2, you do not need to fill in the two items

misss520 commented 1 year ago

I have done all the preparatory work, but when I use Moess for inference, I only get a bunch of noises. The inferred audio can hear the timbre, and the duration is the same, but there is no articulation, similar to a child's mouth. In Webui, my model can be inferred normally, I wonder what else is wrong, I am using SoVits4.0

NaruseMioShirakana commented 1 year ago

If you only got noises, please check your Hubert version. SoVits4.0 has a dedicated Hubert model named "Hubert4.0", you can find it at HuggingFace Place it in the hubert folder, and then change "hubert" to "hubert4.0" in the configuration file.

NaruseMioShirakana commented 1 year ago

If there is still noise, you can try to modifying the "DP噪声规模", Increase its value appropriately

misss520 commented 1 year ago

Thank you, you are a very patient developer, all my problems were solved