Uploaded January 2023 | Updated September 2026, 2 weeks ago

---

MusicLM has taken the Music AI community by storm. The model published by Google is a step ahead towards text-based music generation in the audio realm. By leveraging a clever combination of deep learning base models, MusicLM generates convincing short music clips with good audio fidelity.

Join The Sound of AI Slack Community:
valeriovelardo.com/the-sound-of-ai-community

MusicLM paper:
arxiv.org/abs/2301.11325

MusicLM demo:
google-research.github.io/seanet/musiclm/examples

======================================

Music AI talent recruitment:
thesoundofai.com

Interested in hiring me as a consultant/freelancer?
thesoundofai.com/consulting.html

The Sound of AI Academy:
the-sound-of-ai-academy.teachable.com

Advanced Python Programming:
the-sound-of-ai-academy.teachable.com/p/advanced-python-programming

Connect with Valerio on Linkedin:
linkedin.com/in/valeriovelardo

Follow Valerio on Facebook:
facebook.com/TheSoundOfAI

Follow Valerio on Twitter:
twitter.com/musikalkemist​

======================================

Content:
0:00 Intro
0:45 Text-to-music
2:15 MusicLM demo
3:31 Riffusion and Mubert AI
4:37 MusicLM architecture
6:08 Components overview
7:40 SoundStream
8:20 w2v-BERT
8:51 MuLan
10:51 Training
16:18 Inference
19:52 Experiments
21:52 Limitations
24:40 Thoughts on research procedure