$15–20/hr · Mercor · Hourly, 20 hours a week
A quality expert who listens to AI-narrated French audiobooks and flags technical errors and narration issues to improve text-to-speech models.
What you would do
- Evaluate synthesized French audiobooks for naturalness, accuracy, and quality of the narrated content
- Spot and log technical errors: skipped words, added words, mispronunciations, numbers read incorrectly, and abbreviations expanded wrongly
- Identify unnatural phrasing, intonation problems, and audio glitches that disrupt the listening experience
- Categorize each issue and mark the precise timespan or location where it occurs using the provided tools
- Summarize overall listener satisfaction and describe where narration feels authentic versus where it breaks immersion
Who they want
- Fluent in French (France region) at native or near-native level with genuine audiobook listening experience and strong opinions on narration
- Working English skills enabling clear reporting and team communication on issues identified during review
- Some experience with annotation, transcription, proofreading, linguistics, or detail-oriented review work
- Sharp, patient ear and discipline to maintain accuracy during long listening sessions
- Consistent availability of roughly 20 hours weekly on a part-time schedule
Main skills
What the interview asks about
1.Narration quality and listener engagement
Evaluating naturalness requires not just technical accuracy but understanding how voice delivery affects the reader's immersion and satisfaction.
For example: “Dialogue passage with three characters pronounced correctly but identical intonation for all speakers. How does this break listener engagement?”
2.Pronunciation and linguistic precision
Catching pronunciation and language-specific errors requires deep fluency that goes beyond word recognition to understand authentic French speech patterns.
For example: “An audiobook excerpt mentions a French city, Le Havre. The narration pronounces it with an English R sound instead of the French guttural R. Is this an error you would flag, and why or why not?”
3.Technical error categorization
Distinguishing between different error types helps the AI team understand which aspects of the synthesis need the most improvement.
For example: “In a passage about medical treatments, the text says "10 mg" but the narration says "10 milliliters". How would you categorize this error, and what additional context would you provide?”
4.Sustained attention and consistency
Audiobook QA requires maintaining accuracy across hours of content, and consistency in flagging issues indicates reliability.
For example: “You have been listening for 4 hours and notice that your flagging of minor pronunciation variations has become inconsistent compared to your earlier notes. How would you address this to maintain quality standards?”
5.Context and meaning preservation
Some errors are subtle: the narration is technically correct but fails to convey the author's intended tone or emotional weight.
For example: “A poetic passage uses metaphorical language about silence and absence. The text-to-speech reads it correctly but with flat, monotone delivery that misses the emotional weight. Would you flag this, and as what type of issue?”
A task you may get
Listen to a 30-minute sample of an AI-narrated French audiobook. Flag and categorize all technical errors, narration issues, and areas where listener engagement breaks. Document the exact timestamps, provide examples, and summarize overall quality.
How to prepare
- Listen to 5-10 hours of professional audiobook narration in French to establish your standard for natural delivery
- Compile a personal list of common pronunciation and abbreviation errors you hear in AI narration
- Review the error categorization system used in audiobook QA and practice applying it consistently
- Practice identifying the boundary between minor imperfections and issues significant enough to flag for model retraining
The facts
- Pay
- $15–20/hr
- Hours
- Hourly, 20 hours a week
- Where
- Remote · Remote — preferred: France
- Field
- Language and Audio
- Posted
- 9/14/2026
- Places left
- 10
We wrote this page from the public Mercor listing. It may be out of date, so read the full posting before you apply.