I was paying $5/mo to get access to a high-quality AI voice-cloning model. I had fun for months doing stuff like impersonating Biden in TF2.
Congratulations, team! You've made me proud, and you've made America proud
Now I can't make any more slop because of their new policy.
Liberals really think those stupid facebook fake news memes were the only thing that prevented Hillary from winning. Someone should tell them you can just make up shit like "Biden just made flag-waving illegal" and half the country will believe it; you don't need an uncanny sounding push poll to convince people.
the memes weren't funny, i don't even want to hear a facsimile of his shit-ass voice unless it's reading his will, but go off king.
The mic spamming is whatever but that's an unfair statement because you haven't seen the work of art I never made.
This is cool because anyone who was serious about spreading misinformation, such as dumping huge loads of cash into zucker/musk's ads hole, has no issue spending $10k more for that nvidia setup that small creators can't afford.
Yeah I'm very much a DIY guy but I just don't have the expensive toys for this project. It hurt to pay the $5.
I appreciate "make funny person talk" isn't a huge deal, but as with everything, this will only affect poorer people. Even if every voice gen site ever did this, anyone with a handful of cash to waste can quite easily just get their own voice generation.
The Deprogram (podcast) were using Trump and Biden AI-generated voices to talk to each other and do some bits and hock their merch. I thought it was pretty funny. I wonder if they were using the same service. I guess there are alternatives like Plinky suggested, but yeah pretty lame and hypocritical of the liberal freeze peach worship pretense.
Edit: The very beginning of episode 109 - All Presidents Are War Criminals #2 or #3. You'll never guess who makes a guest appearance and shows his hog.
Damn, we can't get any more bhutanese shadow garden copypastas? Shame.
shit got diamonds on it so you know the CBD to THC ratio is fuckin swag
Everything but my sneakers is AMD.
I avoid nvidia because they don't play nice with Linux.
Could w-okada/voice-changer work? It is one of the AIs I want to set up but I don't know how difficult it can be. I've seen they have an AMD tutorial for linux: https://github.com/w-okada/voice-changer/blob/master/tutorials/tutorial_anaconda_amd_rocm.md
This is a client software for performing real-time voice conversion
Wow! That's impressive. I wasn't looking for real-time processing. That must sacrifice some quality, right?
To be fair, I haven't tried it yet but if it sacrifices quality, you should be able to record the output and then play it back without that loss
I haven't tried yet those cuda wrappers tbh, maybe they work with pytorch
I was paying $5/mo to get access to a high-quality AI voice-cloning model. I had fun for months doing stuff like impersonating Biden in TF2.
Guess you're just going to have to learn to do voices using the old fashioned time tested technique of using your own fucking vocal cords.
Seriously, though, its not that hard to develop a few good impressions. Its a fun little skill and immediately makes you popular at parties. Stop outsourcing your life to a computer. Save your $5/mo. Learn to actually do things for yourself. Also, log off and touch some grass, ffs.
I actually am good at impressions. Fun fact: the meme I was starting to make was going to be a speech-to-speech clone using my own recorded impersonation as a guide for cadence and tonality. The nice thing about that method is that it cleans up the audio so I don't need to use a professional microphone setup to make a decent video production.
fucking hell. imagine thinking that "democracy" is so important yet so fragile that $5 is what separates us from tyranny
