What the work involves
This is a one-off audio contribution, not an ongoing engagement. You receive a short script and a recording guide, read the script aloud in your own voice, and upload the file. The whole task is scoped at roughly 15 minutes including setup. There is no editing, no scripting, and no post-production expected from you — the value is entirely in the authenticity of the speech.
The recordings feed speech datasets used to train and evaluate models on accented English. That means the technical bar is modest but the honesty bar is not: the accent must be genuinely yours. Performed or approximated Bulgarian accents are the single most common reason submissions are rejected, and they are usually audible to a reviewer within a sentence or two.
What the platform screens for
- Genuine linguistic background. Expect questions about where you grew up, what language you speak at home, and how long you have spoken English. These are asked to establish that your accent is native to you, not learned for the task.
- Ability to follow written instructions unsupervised. Nobody will walk you through the recording. The guide covers file format, quiet-room conditions, mic distance and re-take rules, and submissions that ignore it get returned.
- Recording conditions. A phone mic in a quiet room is acceptable. Background traffic, echo-heavy bathrooms, fans, and speakerphone capture are not.
- Turnaround. These projects open and close quickly. Availability inside a short window matters more than total capacity.
Logistics
Fully remote and fully asynchronous — there is no call, no scheduled session, and no supervisor. You work from wherever you are, on your own device. Payment is a flat $50 per accepted submission; if further batches open, contributors who submitted clean audio the first time are typically approached first. Treat this as a small, self-contained task rather than a source of steady hours.