AI-Generated Music (+ Parodies / Covers / Original Content) - The machines are learning art. Observe...

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
I keep seeing example and tutorials everywhere about doing song covers, but I can't seem to find good explanations/tools about text to speech.
I know it exists because we've all heard Emma Watson reading Mein Kampf, what would you say is the best/easiest method to get voices to do a simple text to speech?
Its fallen by the wayside though since using the voice replacement on a recorded audio (same method as music, but no instrumental track) is easier and produces better results.
 
  • Like
Reactions: Claude Sigma
Question for anyone here train their own voice models. How much vram did you guys have? I'm a lolpoor but I compile (ton of ram) and I've got someone willing to loan me several AMD MI8 ML cards but they are Fiji era chips with a 4GB HBM stack.

I shudder at compiling rocm but also wonder if accessing regular RAM is possible, rocm is the sucessor to HSA used in the APUs and rebar (Smart Memory Access). Would be pretty funny to see that CPU memory is also accessible to the GPU.
 
  • Feels
Reactions: Puff
Question for anyone here train their own voice models. How much vram did you guys have? I'm a lolpoor but I compile (ton of ram) and I've got someone willing to loan me several AMD MI8 ML cards but they are Fiji era chips with a 4GB HBM stack.

I shudder at compiling rocm but also wonder if accessing regular RAM is possible, rocm is the sucessor to HSA used in the APUs and rebar (Smart Memory Access). Would be pretty funny to see that CPU memory is also accessible to the GPU.
Mine is 24 GB. If you just want to train RVC I don't think you need a lot, but you'll probably want an Nvidia card.
 
Mine is 24 GB. If you just want to train RVC I don't think you need a lot, but you'll probably want an Nvidia card.
Fwiw I was able to train my Ryuk model on an RX570, which was an 8gb AMD GPU. Took fucking ages to train and live STS isn’t feasible, but the bar’s a lot lower than Stable Diffusion and Dreambooth

If it’s feasible for your situation though, yeah, get an Nvidia GPU. Even the lower end stock can net you decent results
 
Another Rocky one. A bit random
A cover of Bobby Sherman's Julie Do Ya Love Me, but Julie from Scott Pilgrim? Really? How is that possible, how would that even happen?
 
Итан Ральф (Ethan Ralph) - Модная девчонка (1989)
 
Making some of my own.




from random.txt:
 
Glad I found a thread about these, this is like the one genre of ai content that doesn't get old to me when it ends up coming out good.
I don't know if this one's been posted and I'm too lazy to check, but this one is interesting because it's a cover that's got the ai voice put in over the cover. There's a few that do it like this but my bing bing wahoo enjoying self can't help but love this one specifically.

For whatever reason recently a lot of Neco-Arc related ones have been showing up in my recommended section on youtube. A lot of them just end up sounding like pitch shifted mildly nasally version of the original singers but there's been some cases where the iconic "funny raspy loud asian woman voice" of the character actually comes through.
 
  • Islamic Content
Reactions: Catgirl Tyranid
Glad I found a thread about these, this is like the one genre of ai content that doesn't get old to me when it ends up coming out good.
I don't know if this one's been posted and I'm too lazy to check, but this one is interesting because it's a cover that's got the ai voice put in over the cover. There's a few that do it like this but my bing bing wahoo enjoying self can't help but love this one specifically.

For whatever reason recently a lot of Neco-Arc related ones have been showing up in my recommended section on youtube. A lot of them just end up sounding like pitch shifted mildly nasally version of the original singers but there's been some cases where the iconic "funny raspy loud asian woman voice" of the character actually comes through.
What demon possessed shit spawn deemed it necessary to create this disaster? I jumped to a random spot in We Are and was assaulted with the most potent ear rape I've heard in a very long time.
 
What demon possessed shit spawn deemed it necessary to create this disaster? I jumped to a random spot in We Are and was assaulted with the most potent ear rape I've heard in a very long time.
It's literally what the character is supposed to sound like. Shit's funny.
 
Back