Gradio demo app for AudioGen

[ashleykleynhans]

While there is a Gradio app for MusicGen, there does not appear to be one for AudioGen.

Would it be possible to have a demo gradio app for AudioGen too?
Or even better, to combine them both into a single gradio interface?

[JayDeeRC]

A combined version would definitely be amazing. However were you able to get the most recent version of gradio loaded? Mine does not seem to work and can't pull the model.

#160

[rsxdalv]

That's exactly how I did it for my MusicGen UI, here's a Google Colab link
https://colab.research.google.com/github/rsxdalv/tts-generation-webui/blob/main/notebooks/google_collab.ipynb

And here's the repo
https://github.com/rsxdalv/tts-generation-webui

[FurkanGozukara]

That's exactly how I did it for my MusicGen UI, here's a Google Colab link https://colab.research.google.com/github/rsxdalv/tts-generation-webui/blob/main/notebooks/google_collab.ipynb

And here's the repo https://github.com/rsxdalv/tts-generation-webui

i should test this nice

[cocktailpeanut]

I've just sent a PR for this. Instead of trying to make both audio and music work in a unified interface, I just created a separate audiogen_app.py. I figured that the UI may diverge further between audiogen and musicgen since they are for different purposes, so having a separate file might be better until someone figures out that having a single UI is better, without making it confusing for end users. Here's the PR check it out: #185

[bluecamel]

That's exactly how I did it for my MusicGen UI, here's a Google Colab link https://colab.research.google.com/github/rsxdalv/tts-generation-webui/blob/main/notebooks/google_collab.ipynb

And here's the repo https://github.com/rsxdalv/tts-generation-webui

Is there a fix for the dependency conflict? Neither option in the notebook works.

[rsxdalv]

That's exactly how I did it for my MusicGen UI, here's a Google Colab link https://colab.research.google.com/github/rsxdalv/tts-generation-webui/blob/main/notebooks/google_collab.ipynb
And here's the repo https://github.com/rsxdalv/tts-generation-webui

Is there a fix for the dependency conflict? Neither option in the notebook works.

I just tested musicgen small and audiogen medium and it worked. Which dependency conflict are you referring to, is it transformers?

Also, I am running on the Google colabs T4 GPU runtime.

[bluecamel]

That's exactly how I did it for my MusicGen UI, here's a Google Colab link https://colab.research.google.com/github/rsxdalv/tts-generation-webui/blob/main/notebooks/google_collab.ipynb
And here's the repo https://github.com/rsxdalv/tts-generation-webui

Is there a fix for the dependency conflict? Neither option in the notebook works.

I just tested musicgen small and audiogen medium and it worked. Which dependency conflict are you referring to, is it transformers?

Also, I am running on the Google colabs T4 GPU runtime.

Ah, I didn't realize that the colab notebook wasn't set to T4 already. It was failing and pip had complained about the transformers version, so I assumed that was the problem. Thanks!

[GrandaddyShmax]

A combined version would definitely be amazing. However were you able to get the most recent version of gradio loaded? Mine does not seem to work and can't pull the model.

#160

Probably late to the party but here is my version of the AudioGen with MusicGen:
https://colab.research.google.com/github/camenduru/MusicGen-colab/blob/main/MusicGen_ClownOfMadness_plus_colab.ipynb

it also features
AudioGen Model, Multiband Diffusion, Custom Model Support, Generation Metadata and Audio Info tab,
Mono to Stereo, Multiprompt/Prompt Segmentation with Structure Prompts, Video Output Customization
and Music Continuation.

It runs well on the free version of the colab, although the MBD requires much more VRAM if you wanna use it

also a neat ui: