#6868th of 70 in Open modelsFor voice, speech and music
basic-pitch
Python library that transcribes audio to MIDI with pitch bends
- Stars
- 6k
- License
- Apache-2.0
- Last commit
- Nov 2025
- Last release
- Aug 2024
- Language
- Python
4,880 on 20 Apr 20266,045 on 9 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Basic Pitch is a Spotify Audio Intelligence Lab library for automatic music transcription. It runs a lightweight neural network that turns an audio file into polyphonic MIDI with pitch bends, without being tied to one instrument. It ships as a pip package with a CLI and a Python API, and includes TensorFlow, CoreML, TFLite and ONNX versions of the model.
Who it is for: Developers and musicians who need audio-to-MIDI transcription in scripts or apps
Strengths
- Instrument-agnostic polyphonic transcription to MIDI, including pitch bends
- Model ships in TensorFlow, CoreML, TensorFlowLite and ONNX formats
- CLI can also save a WAV rendering, raw NPZ outputs and CSV note events
- Installs with pip and does not require TensorFlow by default
Weaknesses
- Meant for one instrument at a time; multi-instrument mixes are not the target
- Stereo input is down-mixed to mono and resampled to 22050 Hz
- Mac M1 supported only on Python 3.10; README lists Python up to 3.11
- Last release was August 2024
What it needs
- Docker
- Needs librosa, TensorFlow (optional)
- Models: TensorFlow, CoreML, TensorFlow Lite, ONNX
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Quick answers
Is basic-pitch open source?
Yes. Its code is on GitHub under the Apache-2.0 license.
Can you run basic-pitch with Docker?
Yes. Its repository has a Docker setup: a Dockerfile, a Compose file or a published image.
Also in Open models
See all 70| Rank | Project | Score |
|---|---|---|
| 1 | ruviewWiFi CSI sensing on ESP32 for presence, breathing, heart rate and pose | 94 out of 100 |
| 2 | layaSingle-pass multilingual classifier for typed choice, score and yes/no decisions | 88 out of 100 |
| 3 | rf-detrReal-time transformer models for object detection, segmentation and keypoints | 79 out of 100 |
| #4 | timesfmPretrained time-series foundation model for zero-shot forecasting from Google Research | 78 out of 100 |
| #5 | yolov5PyTorch YOLOv5 object detection, segmentation and classification models with training code | 76 out of 100 |