Skip to main content
Back to News Hub
🤖OpenAI
April 25, 2019
General AI

MuseNet

Overview

OpenAI introduced MuseNet, a deep neural network that creates 4-minute musical compositions featuring 10 different instruments. The model merges distinct musical genres, ranging from country to Mozart to the Beatles. Instead of being programmed with rules of music theory, it learned rhythm, style, and harmony by predicting sequential tokens across hundreds of thousands of MIDI files.

Key Takeaways

  • OpenAI has released MuseNet, a deep neural network capable of producing 4-minute musical compositions.

    The system can integrate 10 different instruments into its pieces while blending contrasting artistic styles, such as country, Mozart, and the Beatles.

  • The creators did not explicitly program the network with human music theory.

    Instead, the model discovered inherent patterns of rhythm, style, and harmony on its own by training to predict the next token across hundreds of thousands of MIDI files.

  • employs the same general-purpose unsupervised transformer technology underlying GPT-2.

    This architecture demonstrates that sequence prediction models can process audio data with the same fundamental framework used for text processing.

  • OpenAI created MuseNet to generate 4-minute musical compositions featuring 10 different instruments.

    The network combines diverse musical genres, including country music, Mozart, and the Beatles.

  • The system relies on the same unsupervised transformer architecture used by GPT-2 for sequential token prediction.

OpenAI has released MuseNet, a deep neural network capable of producing 4-minute musical compositions. The system can integrate 10 different instruments into its pieces while blending contrasting artistic styles, such as country, Mozart, and the Beatles. The creators did not explicitly program the network with human music theory.

Instead, the model discovered inherent patterns of rhythm, style, and harmony on its own by training to predict the next token across hundreds of thousands of MIDI files. employs the same general-purpose unsupervised transformer technology underlying GPT-2. This architecture demonstrates that sequence prediction models can process audio data with the same fundamental framework used for text processing.

OpenAI created MuseNet to generate 4-minute musical compositions featuring 10 different instruments. The network combines diverse musical genres, including country music, Mozart, and the Beatles. MuseNet learned rules of rhythm and harmony by predicting the next token across hundreds of thousands of MIDI files.

For more details please read the original article at OpenAI.

Continue Learning

Comments

Comments appear only after moderation. Your email identifies your submission to the moderator and is never displayed here.

No approved comments yet.

Originally published by OpenAI
Read the original