What the work involves

You will work through queues of short Vietnamese audio clips, some spoken by people and some synthesized or generated by a model. For each clip you decide whether it sounds like a native speaker, and where exactly it falls short: a mistimed tone contour, a flattened final particle, vowel quality that drifts between Northern and Southern norms, prosody that scans as read-aloud rather than spoken. Ratings go into a rubric; alongside them you write a short justification in English naming the specific phonetic or discourse feature that drove your score.

The value you add is consistency and specificity. "Sounds off" is not usable feedback; "tone 5 (ngã) realized as tone 3 (hỏi), audible at 0:04 on 'cũng'" is. Expect calibration rounds, occasional disagreement resolution against other annotators, and clarifying threads with project coordinators when guidelines don't cover an edge case. No prior AI experience is expected — the platform is buying your ear and your ability to describe what it hears.

What the screening looks for

  • Verifiable native-level Vietnamese, including comfort discussing regional variation rather than treating one dialect as the only correct one.
  • English strong enough to write evaluative prose — roughly B2 and up. You will be assessed on written clarity, not just comprehension.
  • Phonetic vocabulary or a trained ear: linguistics, phonetics, voice work, audio engineering, transcription, language teaching, or broadcast experience all read well.
  • Rubric discipline — whether you can suppress personal preference and apply a stated standard, and whether you flag ambiguity instead of guessing.

Expect depth follow-ups. Screens on AI-led platforms typically press a second and third time on the same claim, so vague answers about "strong attention to detail" fare badly compared with concrete examples.

Logistics

Remote, contractor, asynchronous. Task batches are claimed from a queue and have deadlines rather than fixed shifts, so most contributors work in blocks around other commitments; many projects want 10–20 hours a week of realistic availability. Audio review requires quiet conditions and decent headphones. Volume is project-driven and can pause or ramp without much notice — treat it as variable contract income, and note the pay band reflects observed rates, not a guarantee.