search

TEAMuP-dev / HARP

A sample editing application allowing for hosted, asynchronous, remote processing of audio with machine learning by routing through Gradio endpoints.
BSD 3-Clause "New" or "Revised" License
51 stars 3 forks source link

proper UI support for generative models that don't require audio input #85

Open hugofloresgarcia opened 10 months ago

hugofloresgarcia commented 5 months ago

@xribene any clue how hard this would be now that we're working with a standalone? @cwitkowitz it would probably involve pyharp API changes right?

cwitkowitz commented 5 months ago

I think it depends on how we want to implement it. Right now, we are moving towards generalizing the input(s) and output(s) for models (there will be a few intermediary steps to this), so once that's finished, there should be no additional API changes. If we want to be able to condition on text, we then either treat text as an additional input (which would then need to be implemented as another type of media display - which could also be loaded with a file), or we can treat text as a control for the model (which shouldn't require any API changes other than support for text controls).