How to Install Meta's FREE Text-to-Music AI Generator Locally (AudioGen)

Показать описание

In this video, I show you how to install and use Meta's new open-sourced music generator, MusicGen. I'll show you how to use it with Google Colab and install it locally using the custom files I created for you. Is this better than MusicLM? Let's find out!

Enjoy :)

Join My Newsletter for Regular AI Updates 👇🏼

My Links 🔗

Media/Sponsorship Inquiries 📈

Links:

Рекомендации по теме

Комментарии

Congratulations on getting this to work on the local computer. After listening to the samples, the short clips do sound decent, but I can't tell for sure how good the complete song will sound overall. Thanks for spending the time to demonstrate this.

marcfruchtman

No updates to this AI music stuff by anyone for over a year now. Sigh.

researchandbuild

Trouble Chute has a one line installer for this that works super easy.

joe_limon

The classic reggae one sounds so familiar to what one would hear from those old adult flash pretty neat

bored

This is awesome Matthew! Thank you for helping bring AI to the masses, as it was meant to be! Keep up the great work! 👍

szghasem

Gave it a try. I did a style conversion of Garou's Theme/I'm a monster. It is a very heavy metal and fast paced song. I converted it to cyberpunk and to epic choir and symphony. I recognized some of the melody but some parts did not translate well. Cyberpunk sounded to happy and bright, and there was no choir. The original song has choir. But did sound kinda good with horns. Its not going to replace musicians or producers anytime soon.

theatheistpaladin

Awesome vid Matthew, I'm an old school hacker myself - the 1st controller I touched was an Atari console my older brother got when I was 9 :D This is turning into a wild ride watching the things Gibson wrote about in my teens turning into real life - your vids have been one of my go-tos over the last couple of months. PS Pipenv also does a pretty good job of sorting things out, but I found that if you arent running in a container you have to get your symlinks and environment variables right and then pip and pipenv can be used really well

WillSteffen

Awesome! Do you think you can make tutorial for installing Udio Al API locally (Udio is another music Al generator)

Reece-ji

For me, the line;

conda install pytorch torchvision torchaudio pytorch-cuda=11.8 rch -c nvidia

Failes because it apparently can't fine certain packages in the current channel?

Rapscallion

Audio 2 audio is amazing at replicating specific style

jopansmark

Is there a music bitrate interpolation model yet? I feel like that's what we really need with these models, half the problem is they sound like some horribly compressed <96kbps audio file.

I'm imagining something that does the audio equivalent of upscaling an image.

AB-wfek

Can't get it installed. When it does the pip install audiocraft it says it needs torch 2.2. But as I watch the install it deletes existing 2.2 and installs 2.1. Chatgpt says to install 2.2. After installing 2, 2 and do the install audiocraft it then says I have 2.2 existing but I need 2.1. Started again with a clean install and it still does the same thing. Giving up.

AI_Image_Master

Is it just me or is the audio-track not aligned with the speaker in this video?

iandanforth

Great overview of this model, i was able to set it up on the first try and it just worked with your guide. I never knew about anaconda but i can already see myself using it for other projects. I'm having a blast creating music clips, even set one as my wake up alarm. I hope they find a way to remove the 30 second duration limit on these tracks but I'm sure that will come with time. Thanks again this was a valuable watch!

satheng