zoom_in Click to enlarge

Google Magenta Performance RNN

▲ ▼

Generator Music Google AI AI Music Generator Model Piano Past Rnn

language

description Google Magenta Performance RNN Overview

Google Magenta’s Performance RNN is a recurrent neural network trained on a large dataset of MIDI music, enabling it to generate new musical sequences exhibiting stylistic characteristics similar to its training data, primarily focusing on piano performance styles.

help Google Magenta Performance RNN FAQ

What kind of music does Google Magenta Performance RNN generate?

Performance RNN generates polyphonic piano performances rather than only single-note melodies. It models expressive timing, note choices, and playing dynamics through MIDI events.

What data was Performance RNN trained on?

The model was trained on the Yamaha e-Piano Competition dataset. That collection contains MIDI recordings of around 1,400 performances by skilled pianists.

Does Performance RNN generate audio directly?

No. It generates MIDI events that specify which notes to play, when to play them, and how hard to play them, after which a synthesizer or sampled piano produces the sound.

How does Performance RNN represent timing and dynamics?

Its event vocabulary includes 10-millisecond time shifts up to one second and 32 velocity events for note intensity. The published representation contains 388 event types, including note-on and note-off events.

swap_horiz

Looking for Google Magenta Performance RNN alternatives? Compare top competitors ranked & scored

arrow_forward