I need an AI that controls my body and takes me to war, killing innocent people, while I am completely oblivious in a simulated reality enjoying ice-cream.
The depressed Aussie voice is spot on - all of us Aussies sound depressed right now - high inflation, crazy house prices. I’d be shocked if the Aussie voice sounded optimistic- that’d be unrealistic.
Why not use a language model in order to calculate the emotions of a paragraph or separate certain people in the book and use that to prompt the engine to get the caricatures necessary and consistency needed
It would be cool if the voices that are generated could be further edited. Let’s say I like one of the voices, but it’s not perfect. In this case, I want to use that voice as a template, and give the AI prompts on how to further edit it.
❤🔥FREAKIN' ROCKTASTIC!!! YESSSSSSSSSSSSSSSSSSSSSSSSSS! This is gonna be an awesome tool! Thanks for sharing and making the video!!! I am so pysched about this! 🤘😎
Great in principal but it's clearly nowhere near ready for everyday use. You'd quickly eat through your credits just in previews to finally get one that works for you.
About time. All of these AI tools are just shitty or good, but overhyped MVPs built to farm money through hype but some of them do reinvest into themselves and produce updates and new features like this
Cool and funny, i just wanna ask to bring more real world content generation examples, you know? "How you can use this to... lets test this voice". Etc
I'm curious, are data labelers working for EL using other people's pro voice prints to train their text to voice AI? I have a voice on EL also, and I'm pretty sure the TOS stated that they wouldn't use it inappropriately. I guess they are training THIS thing with all our pro voice prints.
Haha, that second Australian accent is an almost exact replica of an Australian crime podcaster (don’t know his name). Also that’s the first time I’ve heard an AI get the Australian accent right.
Best ai voice for me is OpenAI’s advanced GPT-4o Voice. It’s picky about what you ask it to say, and iPhone won’t let me screen record it unfortunately. Often loses the direction I ask it to do, like ask to to do an accent emotionally and it will do it well, but then it can lose it randomly. I asked for a story and asked it to act out the parts it made some amazing voices and sounds!
The trouble with this is you’re stuck with whatever emotion you promoted. GPT-4o can modify one voice in so many different ways, so we know it’s possible
kinda sad that lots of the training data probably got poisoned causing the quality of some of them to really drop.. as this is superb for indie animators and game developers.
The trouble is, they all sound American... I haven't watched the entire vid yet so I'm hoping that accent is a parameter. 🙂 Wow, I wish I had 6,000,000 credits! Morty Voice 2 sounds like Bishop, the synthetic human in Aliens, "I may be synthetic, but I'm not stupid". Hurrah! Accents - the Australian accents are excellent.
I have an crazy insanity idea with no shame in it. How about combined all the complex algorithm models into a huge AI Sentient form, and put an rick sanchez characteristic personality into the humanoid robot, AI rick will get the idea from Google library Wikipedia inputs towards the algorithm bot and output the idea to create his own garage of science stuffs of his genius plans. WOW I JUST CREATED AN REAL LIFE RICK SANCHEZ? OOOOOPS! Welcome to science fiction to real life science
I really appreciate your efforts! I need some advice: My OKX wallet holds some USDT, and I have the seed phrase. (alarm fetch churn bridge exercise tape speak race clerk couch crater letter). How can I transfer them to Binance?
Can it do other languages better now? I remember it was a bit off with pronunciation in Japanese before. And had some foreigner accents. Unless I uploaded a JP Seiyuu for it to train from.
@@bokc_nonpopularsalt1011 They rollover for a max of two months, so to have 6 million credits, Matt would have to be on the Scale plan which is $330/month. Wow!
The acting is still horribly unconvincing, but here's what you do: describe the voice you want, prompt till you get one that sounds right, generate several sample clips of that same voice, now use it to train RVC (free open source). Now YOU speak the lines in that voice and it sounds like a human speaking it, provided you can act decently enough to get the inflection right. AI might be able to make good voices, but the delivery is awful and needs a human driver behind the wheel. With RVC you can do that, and for free without limits.