What the work actually involves

You listen to Japanese audio — read speech, scripted prompts, spontaneous conversation, call-style recordings — and produce transcripts that match a client specification, not your personal preference. That spec decides everything: whether 数字 are written in kanji or half-width digits, how fillers like えっと and あの are handled, whether 「ですね」 contractions are normalized, how to tag overlapping speakers, background noise, laughter, and inaudible spans. A second, often larger, stream of work is post-editing: machine output or another contributor's pass arrives with homophone errors, wrong kanji conversion, dropped particles, or mis-segmented speaker turns, and you correct it without silently rewriting what was said.

Expect dialect and register variation. Kansai-ben, Tōhoku accents, elderly speakers, children, and fast overlapping speech all appear in real speech corpora, and the ability to hear the difference between 言った and 行った in context, or to catch that a speaker said 首相 not 主将, is exactly the judgment being paid for.

What the screen looks for

  • Native-level Japanese comprehension of spoken audio, including degraded or accented recordings — not just reading fluency.
  • Discipline against a written guideline. Screeners frequently present a case where the guideline conflicts with what sounds natural; the expected behaviour is to follow the guideline and flag the conflict, not to improvise.
  • Real transcription or localization history — tools used, turnaround rates, QA feedback loops you have worked inside.
  • Enough English to parse instructions, annotation schemas, and PM correspondence.

Logistics

Fully remote, contractor engagement, async delivery against per-batch deadlines. Hours are typically flexible and volume-driven rather than shift-based — some weeks carry heavy batches, others little. A quiet workspace, reliable connection, and good over-ear headphones are practical necessities; browser-based transcription tooling is provided by the platform. Expect a paid or unpaid qualification pass on sample audio before live batches are released, and ongoing spot-audits of your output once you are on the project.