Textually Guided Audio Generation

Text to image stuff is great, but this is what I've been waiting for. There isn't much yet, and the samples are kinda rudimentary, but you know it's coming to take our jobs and its gonna be great for electronic musicians. Just posting this as a hub for this stuff.
https://felixkreuk.github.io/text2audio_arxiv_samples/ 

The talking samples sound like they belong on The Sims.

 

in de col men seivuan, prisencolinensinainciusol, ol rait?

and I gotta say, those DiffSound samples are goddamn nightmare fuel

  On 10/15/2022 at 1:35 AM, luke viia said:

in de col men seivuan, prisencolinensinainciusol, ol rait?

 

and I gotta say, those DiffSound samples are goddamn nightmare fuel

Yeah sounds pretty cool though, I found how to run it here https://github.com/yangdongchao/Text-to-sound-Synthesis -  might give it a whirl