Fix bug with voicecraft dropdowns selecting the first instance of a word
add new noise reduction method that is faster
support flac and mp3 noise reduction
always save wav files as 44100 PCM 16 to speed up LIP and FUZ generation
Bulk RVC now keep sub directories
1.1.7 has been released, this contains a critical bug fix with the LIP generation. It was not resampling correctly, which caused bad LIP files
Fix bad lip generation
Added legacy elevenlabs voices
fixed custom models deletion
fixed custom references not loading
I have released 1.1.5 where I have re-written RVC. It still uses a lot of CPU, but not its much more efficient.
Rewrote RVC to use warm up and cache
This makes it about 85% faster after the first use.
Bulk RVC now goes from 12 hours to 2 hours for protagonist
Switched to latest FFMPEG using SoX for resampling, which is faster and better
Changelog here:
LIP and FUZ creation, with any function or bulk method
Custom Model Support: Train or download a model and you can use it instantly
Bulk RVC: Want to replace the main character? Do it with one simple click, and suddenly Nate now sounds like a super mutant, a synth, or much more. Converts entire folders and optionally creates lips and fuz
Bulk Denoiser: Ever hear a mod that has custom audio, but clearly is lower quality and static? The AI denoiser isolates the audio and upscales the voice at the same time.
Bulk Upscaler: take low quality sound effects and make them sound high def and crisp
Music Generation: Perfect for creating background music or audio jingles. Includes a melody cloning feature using any reference audio. Not trained on any audio... yet. Trained using creative commons sourced training data.
Sound FX Generation: Create Sound effects on demand, trained using creative commons sourced data.
KNOWN ISSUES:
Music Generator isnt releasing GPU memory like it should, need to close the app to clear it.
Sometimes the cache can become corrupted when downloading music, audio, and text models from huggingface. Delete this folder: C:\Users\USERNAME\.cache\huggingface\hub You will see this error in the logs:
ERROR - Error: Model has been downloaded but the SHA256 checksum does not not match. Please retry loading the model.
I have to say it is refreshing to see 5 muntes taken to show how the mod actually works. This is a class mod, only thing im not sure of is how to replace the sound.. i cant for the life of me remember the old mod everyone used to use for making companions etc..
0000B0EF_1 playervoicefemale01 I remember this from the ant colony..you saved a group of initiates with a fatman despite not knowing if they'd die in the blast. 000212a0_1.fuz 0000B031_1 playervoicefemale01 I remember hearing about that memory in the ant colony. Taylor said you saved a group of initiates and that's why you were made Paladin. 000212a0_1.fuz 000418F7_1 playervoicefemale01 Can't you read my mind? 000212a0_1.fuz 000418F5_1 playervoicefemale01 I just didn't like them. 000212a0_1.fuz 000418F3_1 playervoicefemale01 The Institute was the better hope for humanity. Besides my son was with them. 000212a0_1.fuz 000418F8_1 playervoicefemale01 They were a threat to the Synths. 000212a0_1.fuz 000418FA_1 playervoicefemale01 The Commonwealth didn't belong to the Brotherhood, it belonged to the people. 000212a0_1.fuz 000418F1_1 playervoicefemale01 They were evildoers who brought only war to the people of the Commonwealth. I ended that. 000212a0_1.fuz
I can't seem to generate anything for child female. Generate is greyed out so I clicked transcribe and it errors. I also lost all the references when I switched model which sucked, and there's no way to select multiple? Would love some pointers here if you have a moment :D
This is kinda amazing considering all the possibilities... Could this be used to completely change the Protagonist's Voice? For example, there is a Daryl Dixon mod, would you be able to train it to mimic Norman's voice and use that to replace Nate?
wow, this is pretty cool! is there any way to multi-select reference voice lines? I'm assuming selecting more references = better quality but slower generation. Maybe I'm wrong?
*edit: i see my question is answered in the app faq, only one reference needed.
*update: i can't get fuz file generation to work. only wav files are generated. i guess ill use FaceFXWrapper.exe to make lip files and go from there.
184 comments
1.1.7 has been released, this contains a critical bug fix with the LIP generation. It was not resampling correctly, which caused bad LIP files
I have released 1.1.5 where I have re-written RVC. It still uses a lot of CPU, but not its much more efficient.
Changelog here:
KNOWN ISSUES:
ERROR - Error: Model has been downloaded but the SHA256 checksum does not not match. Please retry loading the model.
https://www.nexusmods.com/fallout4/mods/87096
to get this to work I first had to batch generate wav files using GPT_sovits + CSV and a player character voice, then batch convert those with RVC.
I did not see how to batch process directly.
(small typo: also the yellow text in the launcher window says "weclome to falltalk")
0000B0EF_1 playervoicefemale01 I remember this from the ant colony..you saved a group of initiates with a fatman despite not knowing if they'd die in the blast. 000212a0_1.fuz0000B031_1 playervoicefemale01 I remember hearing about that memory in the ant colony. Taylor said you saved a group of initiates and that's why you were made Paladin. 000212a0_1.fuz
000418F7_1 playervoicefemale01 Can't you read my mind? 000212a0_1.fuz
000418F5_1 playervoicefemale01 I just didn't like them. 000212a0_1.fuz
000418F3_1 playervoicefemale01 The Institute was the better hope for humanity. Besides my son was with them. 000212a0_1.fuz
000418F8_1 playervoicefemale01 They were a threat to the Synths. 000212a0_1.fuz
000418FA_1 playervoicefemale01 The Commonwealth didn't belong to the Brotherhood, it belonged to the people. 000212a0_1.fuz
000418F1_1 playervoicefemale01 They were evildoers who brought only war to the people of the Commonwealth. I ended that. 000212a0_1.fuz
https://pastebin.com/r267WQju
*edit: i see my question is answered in the app faq, only one reference needed.
*update: i can't get fuz file generation to work. only wav files are generated. i guess ill use FaceFXWrapper.exe to make lip files and go from there.