What the work involves

This is a single-session speech data task, not an ongoing role. You receive a short script and a written recording guide, record yourself reading the script aloud in your natural speaking voice, and upload the file through AfterQuery's designated submission flow. Start to finish, contributors report the whole thing taking under 15 minutes. There is no editing, no script writing, and no follow-up call — the deliverable is one clean audio file that matches the guide's specifications.

The purpose behind these projects is usually accent coverage for speech recognition and speech-model evaluation. Systems trained mostly on general American or British English degrade on accented speech, so datasets need authentic, natural-sounding samples from speakers whose English is genuinely shaped by Korean phonology. That is why the listing is blunt about imitated accents: a performed accent is worse than useless in a training or eval set, and reviewers listening for it are usually able to hear it.

What the platform screens for

  • Authenticity of accent. The screen is looking for a Korean-influenced English accent that is your actual speaking voice — typically a native or heritage Korean speaker, or someone raised in a Korean-speaking environment. Some screens include a short spoken sample.
  • Instruction compliance. Whether you can read a written guide in English and follow it exactly without asking for hand-holding — file format, ambient noise, distance from the mic, one continuous take.
  • Working equipment. A smartphone or computer microphone is enough; studio gear is not required or expected.
  • Turnaround. These projects open and close quickly. Availability within a day or two matters more than long-term capacity.

Logistics

Fully remote and fully asynchronous — record wherever you are, whenever you like, as long as the room is quiet. Payment is a flat $50 per accepted submission rather than a tracked hourly rate; the $30/hr figure is an observed implied rate derived from typical completion time, and is not guaranteed. Rejected recordings (background noise, wrong format, misread script) may need a redo before payment clears. Repeat invitations to similar audio batches are common but not promised.