Instant Voice Cloning and Speech Editing with Voicecraft

[@tempertephra]

Danke! Great projects for popular usage. ❤I look forward to easily utilize a voice token of german language and want to put the idea on the table, whether it may be more beneficial to have a downloadable database with well trained language tokens and models rather than needing to train this separately for every user. It is easy to get rvc models and so it should be to get language models.🌄

[@tempertephra]

I just saw that tokenizers have a typical size of

[@Jarods_Journey]

Appreciate the super! As for the language models, they're more cumbersome to move around due to being larger, but the biggest issue is just the lack of wide adoption for tortoise. I'm not entirely committed to managing something that's community run, but if there are others that would like the head the task, that'd be great! As for tokenizers, I wouldn't need to make them because the workflow for them before you train a new language, requires it to be already created. There is no one size fits all with BPE tokenizers as it'll vary depending on your dataset.

[@tempertephra]

@@Jarods_Journey Thank you for the update! Seems for now I am loking forward to the float16 fix orrr get an RTX gpu ;) I wonder whetehr there is really no possibility to create a universal language tokens because there seems to be some kind of universal pattern in language. en.wikipedia.org/wiki/Characteristica_universalis

[@vik4741]

Can't wait for the tutorial to host it locally

[@Pacifier1222]

FYI, you need docker desktop with WSL2. After that you can run Google Collab and link your Docker in the web page

[@dougmaisner]

i love the audio stuff. great work. always looking forward to more of your videos and of course the elden ring DLC lol.

[@Jarods_Journey]

Thank you! I took an looking forward to the Elden Ring DLC lOL

[@farsi_vibes_edit]

You should enter the sample sound with a longer duration to improve the quality of the output

[@parzimav]

sorry off topic, is there a good text-speech that can be ran locally, + you can select voices for sample actors/cartoons etc

[@IchWarNivek]

I would like to know it too

[@keithprice3369]

So, it sounds like the Editing version is just for swapping out a word or two, not changing the text completely. Basically for fixing a couple of oopsies. Yes?

[@Jarods_Journey]

I'd say about so, but it was able to do more than a few words

[@Vlad-hm7cj]

do you konw any tts models that I could fine-tune to speak romanian (or any other language)? or have any resources on that? :(

[@tribuzeus]

Only English?

[@mohnishwarker6338]

Hey, thanks. Can I make 2500 words of voice craft in a single audio guide, please!

[@BorygoTomka]

After loading the voice and clicking the last cell, I have a red highlighted line: start, end = get_mask_interval(align_fn, orig_span_save, edit_type) And it doesn't start

[@keithprice3369]

I love what you're doing, but I'm a bit confused. I've watched 3 of your videos where you showcase voice cloning and in all 3, you end up attempting to clone your voice, admitting it was pretty bad, then trying Melina which always sounds great. What's up with that?

[@Jarods_Journey]

Hey Keith, I've made many videos on it, but if I remember correctly, the ones with my voice are demonstrations of the process and not necessarily a demonstration of its quality. The Melina portion and any other good voices are generally the demonstrations of quality to show that it can be very good. Usually the reason it's sounds bad on my voice is because the voice is undertrained. Hopefully this makes sense and clarifies why I did it that way

[@keithprice3369]

@@Jarods_Journey The first video I found from you was from several months ago and was on TurtleTTS. Then I found your StyleTTS, which appeared to be better than Turtle. Now I've found this video on VoiceCraft. Do you feel VoiceCraft is now the best option for TTS voice cloning? If not, which? And can you recommend which of your videos trains on how to get the best clone for the cloning app that gets the best results?

[@StringerBell]

No local install tutorial? :(

[@Jarods_Journey]

Not yet at least!

[@TomiTom1234]

Is there a way to run it locally?

[@Beauty.and.FashionPhotographer]

Can this voicecraft be installed with pinokio onto a mac m2pro ?

[@DorTurkyy]

bro is tortoise updated? female voice never looked exact with me even with 2 H voice samples

[@Jarods_Journey]

The only tortoise update was enabling it for training other languages, if you're training on 2 hours of audio, it comes down to how good your audio dataset is along with if you also trained an RVC model to math pitch

[@StringerBell]

I trained on 48 hours of Bulgarian professionally recorded speech for 100 epoch.. The results were terrible in Tortoise TTS on all epoches.

[@DorTurkyy]

@@StringerBell yeah tortoise is terrible I have rtx 3090 24gb as gpu,Isnt there a better free one?

[@StringerBell]

@@DorTurkyy I'm using RTX 4090 and trained for 22 hours. I have no idea, I cannot find reliable TTS to train on other languages besides English.

[@DorTurkyy]

@@StringerBell bro there's eleven labs but it's paid

[@canberkguitar-sg6qu]

Which one is good for 3min script? For English and Korean

[@sfonetwo]

How long is the tts it can generate

[@planetgamecommunity817]

Adobe VoCo like?

[@zachary3603]

Why'd you delete your local version. Isn't this way more impressive than half the TTS's out there? It sounds way better than even eleven labs half the time xD.

[@ללמד_טבעי]

I have already noticed several tutorials in a row that you talk about complicated and not user-friendly software and services. Why are there no user friendly ways in your tutorials? שמתי לב כבר כמה הדרכות ברצף שאתה מדבר על תוכנות ושירותים מסובכים ולא ידידותיים למשתמש. למה אין דרכים ידידותיים למשתמש בהדרכות שלך?

[@supersonicunitedsupersonic8531]

what about russian language support?

[@M4rt1nX]

so cool. Thanks for this.

[@WorldYuteChronicles]

thanks again

[@blazearmoru]

nice! more ai guides! thank ye!

[@farsi_vibes_edit]

Imagine someone said a sentence that caused him to be found guilty, and then he comes to change his sentence with this software and use it in court.💀

[@jamid_graphics]

Chairman, you said no GPU on thumbnail but just at this moment 1:58 you said for it to run we need a GPU? OH COME ON.....

[@nomikomi]

is a server gpu on the google collab not local gpu