Create a Branch for me please "Extend"

I have made a major update to this project and would like to share it with you for evaluation. https://github.com/Oncorporation/audiocraft is the fork. Updated UI No music length limit Updated export video to be compatible with Facebook sharing Local model ability

#todo create melody guided music that switches to medium model when the melody sample's duration has expired. Currently, it simply loops back through the melody segments.

https://github.com/facebookresearch/audiocraft/assets/5302844/bc873e54-2371-4ca4-beaa-52e18d235136

https://github.com/facebookresearch/audiocraft/assets/5302844/fab00c75-4ecc-430b-a4c5-3ee91037a00a

dmn i liked this game especially mech warrior 3 yep it should be merged

can i forked your work? or we can working together i made webui version and want continuous Improvement。

here is it,only one-click script install now. https://github.com/sdbds/audiocraft-webui

I added it to hugging face also Surn/UnlimitedMusicGen

can i forked your work? or we can working together i made webui version and want continuous Improvement。

here is it,only one-click script install now. https://github.com/sdbds/audiocraft-webui

Awesome, I am still working very quickly and it may be frustrating for you to try to do pull requests until I get everything established

Feel Free to fork! I would love to integrate the one-click script!

I have made a major update to this project and would like to share it with you for evaluation. https://github.com/Oncorporation/audiocraft is the fork. Updated UI No music length limit Updated export video to be compatible with Facebook sharing Local model ability

#todo create melody guided music that switches to medium model when the melody sample's duration has expired. Currently, it simply loops back through the melody segments.

tmpshu7sigv.mp4 tmpfajni_fg.mp4

Yo, these charges are legit. You definitely beat me to the punch lol. I love it. I've tested it out and man. You did an awesome job, especially the UI changes. You can snag my launcher app on my branch if you want. run_app.py in this one. It checks for a venv, creates it if it doesn't exist, installs requirements, and then launches the app.py script. The run.bat file is already licensed under MIT under my concatenator for video editing, so have at it.

I did manage to get above a minute, but cohesion was a problem. I incrementally fixed it up to 40 seconds. You blew way past that lol. I was able to clean the audio generated up though and use it for this before testing yours out.

https://fb.watch/l9YDnfAnJo/

I have been putting most of my effort into [Hugging Face [(https://huggingface.co/spaces/Surn/UnlimitedMusicGen) app as they have given me a GPU grant. Feel free to review there. I am moving updates back to main and Extend branches

At the moment, I have had a big break through in consistency with Melody attached audio files, but it is not perfect. Also, my interrupt function is junk so far. (suggestions welcome)

To get the music more consistent from Melody attached files, I am grabbing both continuation tokens and next segment tokens. They are not mixed directly, some are attributes and some are prompt_tokens.

I can see the fundamental issue due to the lack of a seed, so I used some prompt engineering principals and low randomization to compensate. You cannot generate the final music from either a random new sample of the next segment of music or via generating a continuation of the previously generated music. So, I have a plan!

Working in Tensor math is fun!

Also fixed the mono .wav output issue