The work

You receive batches of short Turkish audio clips and score them against a rubric: nativeness, fluency, pronunciation accuracy, intonation and prosody, and overall linguistic authenticity. Some clips are human speakers; some are model output being checked for whether it passes as native Turkish. For each rating you write a short justification in English naming the specific feature that drove the score — a devoiced final consonant, a misplaced vowel harmony, flat sentence-final intonation, a stress pattern borrowed from English, a regional realisation that is authentic but not the register the guideline asks for. Vague notes like "sounds unnatural" are the most common reason work gets sent back.

Expect to flag edge cases rather than force them: heavily accented but genuinely native regional speech, code-switched fragments, clipped or noisy recordings, clips where the audio quality makes the phonetic judgement unsafe. Coordinators want ambiguities raised, not silently resolved. Calibration usually starts with a small paid or unpaid sample set and periodic agreement checks against other annotators.

What the screen looks for

micro1's screening is AI-led and largely conversational, with a spoken component in Turkish and written responses in English. It is checking three things: that your Turkish is genuinely native-level rather than advanced-learner, that your English is comfortably B2 or above in writing, and that you can describe a sound difference in concrete phonetic or descriptive terms instead of impressions. A background in linguistics, phonetics, voice work, audio engineering, teaching, or media is preferred and will be probed for specifics — but no AI experience is required.

Logistics

  • Fully remote, contractor engagement, asynchronous task queues with deadlines rather than fixed shifts.
  • Volume fluctuates with client demand; treat it as supplementary rather than guaranteed full-time hours.
  • You need quiet listening conditions and decent closed-back or over-ear headphones — laptop speakers are not sufficient for judging pronunciation detail.
  • Pay of $30–65/hr reflects observed rates on this project and varies by specialisation and task type.