Suggestion

[tomas122727]

Could you also try to add the old thinker anonymous text to speech tool I will hmget up later today and send you a link to it including the models

[hydrusbeta]

Hello tomas122727,

Do you mean tkinterAnon's TTS tool? I saw a video of it being used once but I have not played around with it myself. Hay Say is in desperate need of more TTS options, so I would definitely like to add it if possible. I look forward to receiving links to the tool and models from you.

Right now, I am working on adding RVC, and after that I would like to tackle the problem of Hay Say taking up so much hard drive space, so be aware that it may take a while for me to get around to adding TkinterAnon's tool.

Best, HydrusBeta

[tomas122727]

yes sorry it was late when i typed that

[tomas122727]

Here is link to models (reminder Some Links Do Not work Anymore For Some Reason ask epople on PPP forumn to reuplaod them if possible and update links but some still work https://docs.google.com/document/d/17VAnMQI4NJzu7UXZALs14AFvhpw8wvbLdA9HrA2xLus/edit

sending u a link to tksynthesis soon it also needs major updates and fixing i wonder if your able to do that possibly cause it stopped launching on my pc for some reason

[tomas122727]

https://mega.nz/file/O5R1yDgA#CbfceC-yk_X7-AbF8xIQz3QDHbg6w-35bHoSUrfd3Ig new link to tksynthesis

[tomas122727]

tho atm i cant find the patch now im gonna see if i have that on my hardrive ill send it right away

[tomas122727]

ur in luck i do :D

[tomas122727]

Link to The Patch https://mega.nz/file/61Z3lRQC#zGOHT1Rv7yKHcpwo6xXGxDy-0QlPsSU70YbG9yl6s7o

[tomas122727]

you could also use tortoise tts include the training tabs and all that so ppl can train their models as normal on their to

[hydrusbeta]

I took a look at Tortoise about a month ago and managed to train a Rainbow Dash model. Here is some sample output: https://files.catbox.moe/is2y8a.mp3 I'll plan on adding Tortoise to Hay Say in the near future, although it will only be for inference, not training. Using Hay Say for training is tricky because the user would need to load all of the audio training files into the docker container first and it is difficult to view or manipulate the files once they are loaded in. It would take a fair amount of work to make it more user-friendly and my guess is that not very many people would use it, so I'll focus on an inference UI.

Thanks for the links to tksynthesis, the patch file, and all the old Tacotron models. Although they're all crossed out in the doc, it seems that some of those models still exist.