What the work actually involves

You receive short clips — often under a minute — paired with model-generated descriptions or transcripts. Your job is to watch and listen closely, then decide whether the description matches what is actually on screen and in the audio. That means catching hallucinated objects, missed speakers, wrong action verbs, overlooked on-screen text, misattributed sounds, and timing errors where an event is described as happening at the wrong point in the clip. Every rejection needs a short written justification that cites the specific moment and the specific guideline clause it violates. Volume matters, but so does consistency: the same clip type reviewed on Monday and Friday should get the same verdict.

A significant share of the day is guideline interpretation rather than pure observation. Project protocols change, edge cases accumulate, and you will be expected to flag ambiguity upward to project coordinators rather than quietly invent a rule. micro1 runs an AI-led screening interview before matching, so expect to be asked about your reasoning live, with follow-up questions that push on borderline cases.

What the screen looks for

  • Concrete annotation history — tool names, project types, clip volumes, QA scores if you have them. "Content moderation" or "transcription" counts; say so plainly.
  • Written English under pressure. Feedback is the product here. Vague notes like "description inaccurate" fail the quality bar.
  • Calibration. Can you separate a genuine error from a stylistic choice you personally dislike? Over-flagging is as costly as under-flagging.
  • Reliability in async work — stable internet, headphones, a quiet enough environment to hear audio detail, and honest hours per week.

Logistics

Fully remote and contractor-based. Hours are typically flexible within project deadlines, and batch work arrives in waves rather than a steady daily stream — some weeks are heavy, some are quiet. No prior AI experience is required, and the listing says so explicitly; annotation discipline and English clarity carry more weight than machine-learning background. The $8–10/hr band reflects what candidates have reported for this listing and is not a guarantee.