AI Training · Independent contractor
Identify instruments and instrument families in short audio segments to improve music source separation. Professional music experience is required.
THE OPPORTUNITY
Identify instruments and instrument families in short audio segments to improve music source separation. Professional music experience is required.
MUSIC PROFESSIONAL - Stem Source Separation Expert - Proof of Experience Required
Current open-source source separation models (HTDemucs, Open-Unmix, MDX-Net, RS-RoFormer family) separate mixtures into only 4–6 stems. Real music mixtures typically contain dozens of distinct instruments, so this coarse granularity limits downstream applications like remixing, music education, transcription, and music information retrieval. This project collects annotations to extend SOTA source separation models beyond the standard 4-stem paradigm by having annotators identify specific instruments and instrument families present in short audio segments.
Fees vary based on service and complexity
Requirements you need: Location match
Role details are copied from the Meridial listing and can change. Check the platform before you apply.
KEEP EXPLORING
Related roles across other platforms, selected by field, title and skills.