What the work actually involves

You listen to short audio clips — often isolated notes, phrases, or brief passages — and make two calls on each: the instrument family (strings, brass, woodwind, percussion, keyboard) and the specific instrument within it. Turing supplies reference audio and an annotation guideline document; your job is to apply them consistently rather than improvise your own taxonomy. Expect edge cases that make the difference between a trained ear and a casual listener: viola versus violin in the overlapping register, oboe versus cor anglais, upright bass arco versus cello, synthesised emulations of acoustic instruments, and clips degraded by compression or noise. Throughput matters, but so does not guessing — how you handle genuine ambiguity is typically what separates annotators who get extended from those who don't.

What the platform screens for

There is a roughly 15-minute qualifying assessment before onboarding, and it is a listening test as much as a form. The screen is checking that your formal music background is real and specific — degree, instrument, where you studied, what the last three years of paid work actually looked like — and that you can describe why you hear an instrument the way you do, in timbral terms rather than vibes. Expect follow-ups that push on a claim you make. Stating that you can distinguish a clarinet from a bass clarinet invites a question about register break and reed noise; a vague answer at that point costs more than never having made the claim.

Logistics

  • Fully remote, paid in USD, asynchronous within the project window
  • Two-day window with flexible hours; the ~$150–$160 figure is observed for completing a full push across both days, not a guaranteed rate
  • Pay period is unconfirmed on the platform listing — clarify before committing
  • Possible extension to further audio work for strong, consistent annotators
  • You will need decent monitoring: reference headphones or studio monitors and a quiet room. Laptop speakers will cost you accuracy on low-register and timbral-edge cases.