What the work involves

You receive batches of Korean audio and produce verbatim or lightly cleaned transcripts according to a written style guide that will change as the project matures. Most files contain natural, unscripted speech: overlapping speakers, dialect and regional pronunciation, filler and hesitation, background noise, and frequent code-switching into English. A second stream of work is review — auditing other contributors' transcripts against the audio, correcting errors, and flagging systematic problems rather than silently patching them.

Alongside the transcript itself, you add metadata and contextual notes: speaker turns, inaudible spans, non-speech events, register or dialect observations, anything that helps a model learn the difference between what was said and what was meant. Coordinators expect you to ask questions when the guide is ambiguous, and to document your reasoning when you make a judgment call.

What the platform screens for

  • Verifiable Korean depth. Native or near-native fluency, with confident handling of spoken registers — honorific levels, casual speech, regional variation — not just standard written Korean.
  • Transcription discipline. Prior paid transcription work, ideally where accuracy was audited. Expect questions about your typing throughput, your timestamping habits, and how you treat disfluencies.
  • Guideline literalism. The most common failure is a transcriber who "improves" the speech. Screens probe whether you can follow a convention you personally disagree with, and raise the disagreement through the right channel.
  • Written English. You'll be documenting ambiguities and giving process feedback in English.

Logistics

Remote, contractor, asynchronous. Work is delivered in batches against deadlines rather than fixed shifts, so you set your own hours — but batches are throughput-sensitive, and consistent weekly availability matters more than a large one-off burst. Pay is stated as observed for this listing, not guaranteed; rate and volume typically depend on whether you're doing first-pass transcription or review. No AI background is required.