Fretiq identifies guitar strings in browser with 97% accuracy

Identifying which string produces a given pitch on a multi-string instrument like electric guitar is a classic audio challenge—one pitch can often be played on multiple strings at different fret positions, with timbral differences imperceptible to most humans. Aadi Garg's latest paper presents Fretiq, a preliminary browser-native string classification system that tackles this using engineered spectral features and Mel-Frequency Cepstral Coefficients (MFCCs). The system uses a 26-dimensional feature vector (frequency band energies, spectral statistics, 13 MFCCs) and achieves 97.1% shuffled frame-level validation accuracy across 322,215 balanced monophonic frames. An ablation study confirms MFCCs are the primary driver, boosting accuracy from 92.2% to 97.1%.

Fretiq also introduces Comparison Training—a deliberate data-collection methodology where adjacent open-string and fifth-fret pairs are recorded in alternation. This reduces the D3-to-A2 frame-level confusion rate by 44%, though results on other target pairs were mixed. In a more realistic held-out free-play evaluation on 103,000 frames, overall accuracy reached 87.8%. The feature extraction pipeline is implemented in both Python and TypeScript to guarantee training-inference parity, and the paper documents two critical implementation failure modes. Crucially, Fretiq runs entirely in the browser with no hexaphonic pickup, fretboard sensor, camera, or multi-microphone setup—making it a practical, zero-hardware solution for guitarists, educators, and audio developers.