The work
You listen to text-to-speech audiobooks in French the way a paying listener would — but you keep a hand on the timeline. When the narrator reads « 1789 » as un sept huit neuf, when a liaison is forced where it shouldn't be, when St-Étienne comes out as Saint instead of the place name, when the intonation of a question lands flat, you stop, mark the span, choose a category, and write a short note in English explaining what went wrong. Some sessions are error-hunting; others ask for a holistic judgement — was this chapter actually enjoyable to listen to, and if not, where did the experience break?
The volume matters as much as the precision. The models improve from consistently categorized, tightly bounded spans across many hours of audio — not from a handful of brilliant catches surrounded by drift. Roughly four hours a day, five days a week, is the expected rhythm.
What the screen looks for
Mercor's screening is AI-led and conversational. Expect it to test three things: that your French is genuinely native or near-native to the France variety (not Québécois, not Belgian, not learned-abroad), that you are a real audiobook listener with opinions rather than someone who has heard of audiobooks, and that you can describe a speech error precisely enough that an engineer could act on it. Vague answers get follow-ups. Saying "the pronunciation sounded off" invites the question of which phoneme, in which word, and what should it have been — have an answer ready.
Logistics
- Remote, anywhere, though living in a francophone region is preferred
- Part-time, ~20 hrs/week, largely async within loose daily expectations
- Observed pay band of $15–20/hr; rates on Mercor vary by project and are not guaranteed
- Working English required for reports and team communication — you do not need to be bilingual
- Prior annotation, transcription, proofreading, or linguistics work is expected in some form