Комментарии:
IMPORTANT NOTE PLS READ The model on google colab is only using half performance due to restrictions of the free google colab. A Local install on a more powerful GPU should produce better results!!!!
ОтветитьIt doesn't seem better than what I have tried so far, but it's pretty good. If it's really free to use, I will give it a try.
ОтветитьGuess I can cancel my ElevenLabs API subscription :D
Ответитьcan anyone point me twords a local install tatorial?
ОтветитьAre fans of this channel called MVP's?
Ответить"Trust because you are willing to accept the risk, not because it's safe or certain." _Motivation
ОтветитьGrow your video for my subtitle services in very cheap price so contact me
ОтветитьIt's very impressive. Unfortunately in the fine details at the bottom it says it's not available for commercial use.
ОтветитьDude, demos are very exaggerated. Wake up
ОтветитьIt. It now working on Google colab voice not changing can any one help
ОтветитьHow do I do this? Do I have to have any computer programming knowledge to do it? Or is there a website. This is really cool!
ОтветитьA moment of silence for everyone that isn't subscribed to MattVidPro.....always keeping us up2date!!!
ОтветитьYes its interesting, but that means WE CAN'T TRUST NO SPOKEN WORDS! !!!
ОтветитьMorgan Freeman is going to be narrating a lot of short films now.
ОтветитьCanadian sounded very robotic.
ОтветитьI have hours of recording of deceased relatives. Hopefully a longer training set or data set will yield much more fidelity with a lot more nuance, because it’s only greater sampling that you really get to sense the idiosyncrasies of a person. When will that be available?
ОтветитьGreat post. Agreed. Ai models would greatly benefit from open sourcing more.
ОтветитьDoes it work using AMD GPU and Windows?
ОтветитьIt's so bad... This is not in a good level
Ответитьok hear me out
they should use this as a way to voice dva sfm's 🔥
The french DVa had a little accent but that was pretty good. And usually french is bad so that's nice to hear a good generation
Ответитьit's just about time and the best free things come
ОтветитьCoqui ❤
Ответитьthe demo was a catfish lol
ОтветитьHow would this work for musicians who are not singers, but have their own original music that they need vocals for? Would this work like something along the lines of Emvoice..?
ОтветитьImportant to note, that it is "Source Available" and not "Open Source" (As per their License, one can not use it commercially).
ОтветитьHow can you say this is freakishly good if you admit that the Demo is way better than the product itself?
ОтветитьThis is crazzzzyyy good.
Ответить"Read me a bedtime story, Elon. Do all the voices."
ОтветитьI've tested the colab one, and it was quiet terrible. Nothing like my voice at all. Hope the full version is more usable.
ОтветитьOmg 😅this is so cool!! Thanks for sharing this with us ❤
ОтветитьWould love to see this ran locally on gpu without the limitations though.. I can't say it's the best free open source voice cloning.. Not by a long shot, but it's pretty good and it can do text2voice unlike Okada though..
I'm gonna check when I have the time how to install this locally, I could use it along Okada for different purposes
Yay, a pile of code and other vomit on github, which no normal people can use, yay! Awesome. Next?
ОтветитьI'm sure the cloning of political people is why Obama didn't work well
ОтветитьI wonder if you run this by Adobe Podcast afterwards, it'll sound more natural!
Ответитьthe 3,2,1 was chromie from Warcraft :D , great video Matt !
ОтветитьHi Matt! Again a very interesting video. But how can I use the this "jupyter notebook", because I need german language. Cheers
ОтветитьNon commercial license 😞
ОтветитьOne of the big trends in AI at the moment are examples that are cherry-picked, and cannot easily be replicated. Maybe a new, more believable, standard could be for AI companies to ask an outside agency to produce the examples of their products.
ОтветитьIt did not do very well with your voice...
ОтветитьReally interesting, I compared it locally to CoquiTTS and it seems that it's way better for some voices, and way worse for others. On average it's on par with it but in a very complementary way. I suppose the way it works is they start from a reference voice and modify it to reduce the difference gap between it and the input so, if their reference voice is already pretty close it works wonder, but the opposite is also true.
ОтветитьAnyone getting vibes of Terminator ability to clone voices of his victims?
ОтветитьAs a french speaker, I can say that the french speaking is a little robotic, like someone unnaturally exaggerating articulation to avoid any sound blending between words.
Ответитьits no better than any other free open source voice cloner ie total crap!
ОтветитьGerman is ok, but has a american/strange accent
ОтветитьThe license is Creative Commons Attribution-NonCommercial 4.0, so the use case for this code is limited.
ОтветитьWhen they use the idiot Musk to demonstrate their products its simply a no-no for me. Why are people sucking up to that douchebag still? Also the thing doesn't sound anything like you so total fail.
ОтветитьVan use it for audiobookd
ОтветитьSouth African here, and yeah its pretty good! Not perfect, but almost there, first time I heard a SA accent with AI LOL
ОтветитьHey, Matt. In my experience as a seasoned prompt engineer, the crux of the issue lies in the misalignment of NPUs during tensor randomizations in the preliminary tests. It's akin to attempting vector factoring with different tools – without a standardized NPU, the subtle nuances crucial for optimal voice cloning get lost in the mix. Achieving consistency in NPU specifications is paramount to unlock the full potential of this technology, and it's evident why the results in the video may have fallen short of expectations.
I really enjoyed your excitement, though. As always, I'm giving you a thumbs up for your hard work.