Chordmini is an ONNX export of the ChordMini chord recognizer, described as ChordNet (2E1D), designed for chord recognition in music information retrieval contexts. It functions as a classifier that processes log-constant-Q transform (log-CQT) feature windows and returns per-frame chord logits, supporting a large vocabulary with 170 chord classes. The model is intended to be used with the musetric packages/ai runtime and is compatible with onnxruntime-web, specifically leveraging WebGPU for efficient processing.
Feature extraction is not included within the model itself; instead, the host environment is responsible for computing a recursive constant-Q transform on WebGPU and providing the resulting features as a GPU buffer to the model. This separation means that Chordmini is not a direct audio-to-chord solution but rather expects pre-processed log-CQT features as input. The audio input should be mono PCM at 22050 Hz, with an arithmetic-mean downmix applied, before undergoing the CQT transformation.
Chordmini is distributed under the MIT license and is available via Hugging Face. Its design is tailored for integration into music information retrieval workflows where GPU-accelerated feature extraction and chord classification are required.
Chordmini is a Voice, TTS & speech product. It enables developers to perform automatic chord recognition on audio files for music analysis and retrieval. It is built as an open-source project for music information retrieval researchers and developers. Chordmini is open source under the MIT license. The product ships for the web and the command line, and it can be self-hosted.
musetric builds and maintains Chordmini, and the product first shipped in 2020. Development happens publicly on GitHub with 231 commits in the last 90 days. Key capabilities include chord recognition, ONNX export, and webGPU support.
Latest indexed changes and source events
musetric/chordmini-onnx verified by the PulseGate indexer
Other apps tracked under the same category.
