Riffusion - AI generates audio from text

[BertKoor]

Generate audio using the Riffusion model.
In single prompt mode you can generate up to ~1 minute of audio with smooth transitions between sections. (beta)
Bi-prompt mode interpolates between two prompts. It can generate up to ~2 minutes of audio, but transitions between sections are more abrupt.

https://huggingface.co/spaces/anzorq/riffusion-demo

[BertKoor]

An in-depth article about consequences of AI:
https://danielbmarkham.com/the-overlord ... showed-up/

[greententacle]

An in-depth article about consequences of AI:
https://danielbmarkham.com/the-overlord ... showed-up/

nice one, thanks!

[SampleScience]

A.I. is not as efficient with audio as it is with graphics and text. But I'm sure it will come to a point where it'll be as good, maybe in a few years? or months?

[tatanka]

For this approach I would definitely prefer this app:

https://www.audiocipher.com/?utm_campai ... 285528e699

It is version two now with chords generation added to melody generation and the guy who develloped it has a great overview on all whats going on in the AI scen and is posting regularly great infos about all this stuff with fair classifying his own app.

Of course its far away from something like Wotja but its just fun and you dont have to make a "university study" to get into it

[Srae]

A.I. is not as efficient with audio as it is with graphics and text. But I'm sure it will come to a point where it'll be as good, maybe in a few years? or months?

Have you heard what soundful and soundraw can generate? They caught up pretty fast..

[SampleScience]

A.I. is not as efficient with audio as it is with graphics and text. But I'm sure it will come to a point where it'll be as good, maybe in a few years? or months?

Have you heard what soundful and soundraw can generate? They caught up pretty fast..

I gave them a try, and I'm not impressed. Their licensing system shows that you can't do much with this kind of technology applied to music. At least, with graphics A.I., you can already commercialize your outputs. You can even sell your prompts. Soundful and Soundraw seem to be based on their own limited music library, so the results are weak and you don't have nice surprises.

The result why music A.I. is so bad right now is because the copyright laws for music are enforced. The copyright laws for graphics are not as strong, hence Midjourney, Stable Diffusion, Dalle, etc.

[osiris]

Unless it can Save to Midi, I can't use it.