Job Description

This role involves contributing real music recordings paired with precise MIDI annotations for an AI training dataset. The project requires musicians, producers, or audio engineers who can supply existing or newly recorded tracks featuring exactly three or four instruments, including vocals if desired. The focus is on genres such as Jazz & Blues, Electronic & Hip Hop, Folk/Singer-songwriter, Heavy Metal & Rock, and Contemporary & Classical. Candidates with finished tracks and available session files are especially encouraged to apply, as minor edits to fit the requirements are acceptable. The work is ideal for those with access to a DAW and the ability to export and correct MIDI files to match audio recordings with high accuracy.

Deliverables

• Provide one real audio recording (3–4 minutes) featuring exactly three or four instruments per track.
• Submit two MIDI files per song: one as captured by your DAW and one with every note ending corrected to match the audio precisely (within 10–20 ms).
• Ensure at least one instrument is genuinely performed; a mix of performed and programmed parts is acceptable.
• Include sung vocals if desired; rap or spoken vocals are not suitable for note-level annotation.
• Deliver a lossless WAV file (44.1 kHz or higher), both MIDI files, and a completed metadata sheet for each song.
• Ensure all audio and MIDI are captured in the same take, from existing DAW archives, or via manual transcription with hand-checked accuracy.
• Follow provided guidance and deliver within one week of song approval.