What the work involves

This is applied audio and language work rather than general annotation. On a typical shift you might listen through Hindi recordings and flag clipping, background noise, mic handling artifacts, or channel imbalance; verify that a model-generated transcript matches what was actually said, including code-switched English tokens and regional pronunciation; or rate synthesized Hindi speech on naturalness, prosody, and word-level intelligibility. Some projects ask you to segment long recordings, mark speaker turns and overlaps, or write short diagnostic notes explaining why a clip fails spec.

Other streams involve producing rather than reviewing: recording scripted or spontaneous Hindi in a controlled environment, checking your own takes against loudness and noise-floor targets, or building small reference sets that downstream annotators are calibrated against. Expect written guidelines that change between batches, and expect to be measured on whether you follow the current version rather than your own preference.

What the platform screens for

  • Native or near-native Hindi, including comfort with Devanagari orthography and with the Hindi–English mixing common in natural speech.
  • Real audio literacy — you should be able to talk about sample rates, bit depth, LUFS and dBFS, noise floor, clipping, and common DAW or tooling workflows without hedging.
  • Consistency under follow-up. micro1's screen is AI-led and asks probing follow-ups; vague answers about "checking quality" tend not to survive them.
  • Written English sufficient to document a defect clearly for an engineering audience.

Logistics

Remote and largely asynchronous, with work delivered in batches. Contributors commonly commit 10–25 hours per week, though volume fluctuates with client demand and can pause between projects. You will need quiet listening conditions, decent closed-back headphones, and for recording projects a usable interface and mic. Pay is stated as observed on the platform, not guaranteed, and varies by project complexity and QA tier.