CREPE
CNN-based monophonic pitch (f0) tracker with state-of-the-art accuracy.
| Language | Python |
|---|---|
| Category | Audio analysis |
| License | MIT |
| Platforms | macOS Windows Linux |
| Install | pip install crepe |
| First released | 2018 |
| Maintained | Not actively maintained (last activity 2024) |
| Links | Website · Source |
Strengths
- Highly accurate monophonic pitch estimation
- Per-frame confidence output and optional Viterbi smoothing
- Simple CLI and Python API
Limitations
- Monophonic only; not suitable for chords/polyphony
- Ageing TensorFlow dependency, last updated 2024
Best for
Precise fundamental-frequency tracking of single-voice audio (vocals, solo instruments).
Minimal example
import crepe
from scipy.io import wavfile
sr, audio = wavfile.read("audio.wav")
time, freq, conf, act = crepe.predict(audio, sr, viterbi=True)