Audio samples

[andimarafioti]

Do you have some audio samples to listen to?

[Njasa2k]

+1

[painebenjamin]

I'll let the authors comment on whether or not something is incorrect, but this is the output I get when running my fork that removes hardcoded paths and downloads models from HuggingFace using the seed and prompts provided, and the giant model, then converted to MP4 so it plays nice in GitHub.

python sample.py --version giant --prompt_file config/example.txt

The song is an epic blend of space-rock, rock, and post-rock genres.

sample_19.1.mp4

This song is a mashup of punk, progressive, rock, and no wave genres.

sample_4.mp4

This song is a fusion of both heavier metal and more melodic rock genres.

sample_8.mp4

This song is a classic rock anthem with driving guitars and powerful lyrics.

sample_17.mp4

All the samples from the example file: https://github.com/painebenjamin/FluxMusic/tree/main/wav

[feizc]

Hi, there is no problem with code inference :)
As the currently provided ckpts trained 200K steps for scaling ablation with subset of training.
We will provide the full version and gradio demo as soon as possible.
I'm very sorry for the misunderstanding.

[UstymKhaburskyi]

Hey @feizc !

Thank you so much for sharing the code and the detailed paper! Could you possibly share a few examples of the final model’s output? The ones trained for just 200k steps don’t seem to fully showcase its potential, and I’m really curious to see what the model is truly capable of. Thanks again!