What the work actually involves
You listen to short clips — often a few seconds — and make two calls: the instrument family (strings, brass, woodwind, percussion, keyboard) and the specific instrument inside it. Reference audio is supplied so that your judgments anchor to the project's definitions rather than your own, which matters when the distinction is viola versus violin, oboe versus cor anglais, or a sampled Rhodes versus a physical one. The annotation guidelines are the source of truth; where your ear and the guideline disagree, the guideline wins and the disagreement goes in a flag or comment. Volume matters here — this is a throughput task with a quality gate, not a slow analytical one.
What the screen is looking for
A ~15-minute qualifying assessment precedes onboarding, and it is an ear test in practice: family classification first, then specific instrument, on clips that include timbrally adjacent pairs and processed or low-fidelity recordings. Turing screens for a music degree plus three or more years of real performance, production, or engineering work, and for the temperament that survives repetitive listening without drifting into guesswork. Prior exposure to transcription, remixing, music information retrieval, or data annotation is a genuine advantage rather than boilerplate — it usually means you already think in terms of label taxonomies and edge cases.
Logistics
- Fully remote, paid in USD, async within a fixed two-day project window
- Observed pay is roughly $150–160 for a full push across both days; the period and the per-task structure are not confirmed by the platform
- Flexible hours inside the window, but the window itself is short — this is not a trickle-in-when-you-can project
- Extension onto further audio work is possible for annotators whose agreement rates hold up
Use decent closed-back headphones or monitors. Laptop speakers will cost you on percussion and low-register woodwind distinctions, and the quality gate will notice before you do.