Musical Deep Neural Networks in the Browser
summary
Neural music models run entirely in the browser using TensorFlow.js and Magenta.js, loading pre-trained nets (roughly 20MB each) to generate melodies, drums and multitrack loops. Examples include an arpeggiator powered by improv_rnn, a drum machine using drums_rnn, and a 7x7 interpolation grid driven by MusicVAE that explores a ~256‑dim latent space. Code uses MIDI‑like quantized seeds, temperature controls, encode/decode and interpolation methods; demos load, generate and sound as promised in real time.
AI-written summary · about
chapters (12)
- 0:00Introduction to generative AI
- 1:05Abstractions in browser AI
- 2:29Arpeggiator demo with Magenta
- 4:03How the melody model works
- 5:46Drum machine with Magenta
- 6:59Drum RNN API details
- 8:03Performance RNN explained
- 8:48Musical spinners from latent space
- 11:19Music VAE geometric space
- 13:52VAE interpolation in code
- 15:37Magenta models and APIs
- 16:18Q&A session
related talks
28mMusic and AI in the Browser with TensorFlow js and Magenta js
24mCreating an AI Musician with JavaScript
41mMusic and Machine Learning
36mMusic and Art with ML
46mMusic Generation with Magenta: Using Machine Learning in
47mExpressive AI-Generated Music With Google's Performance RNN
01h 04mKeynote 3. Interactive generative models for music
36mMachine Learning in the Browser