search

Stability-AI / stable-audio-tools

Generative models for conditional audio generation
MIT License
2.25k stars 197 forks source link

Multi-channel data support #94

Open fmiotello opened 3 weeks ago

fmiotello commented 3 weeks ago

It seems that audio data is forced to be 1 or 2 channels, isn't it? Is there a reason for this? Is there a way to simply enable multi-channel (>2) data support?

fred-dev commented 3 weeks ago

It seems that audio data is forced to be 1 or 2 channels, isn't it?

Is there a reason for this? Is there a way to simply enable multi-channel (>2) data support?

The stereo here is quite an achievement. I am sure someone is working on further multichannel support but it is not as easy as adding channels. Take a look at the paper for more details: http://www.arxiv.org/abs/2402.04825