Training Turk

Mercor listing

Voice Actor: CX Agent Voice Cloning (South African English)

$50–100/hr

The role in one line

Voice actor with South African English accent who records voice samples for AI text-to-speech and voice cloning.

Written by Training Turk from the public listing; it may be incomplete or out of date. Read the full posting on Mercor.

What you would do

  • Capture voice recordings in various styles - conversational, narrative, instructional - with high audio quality
  • Produce clear, lifelike, emotionally nuanced speech with refined command of vocal inflection, timing, and articulation
  • Perform multiple takes of the same text with variation in emotion, emphasis, and vocal style
  • Keep your voice tone, regional accent, and delivery style uniform throughout multiple sessions
  • Follow detailed recording guidelines on microphone setup, environment, file formatting

Who they are looking for

  • Native English speaker residing in South Africa with a genuine South African accent
  • Clear, natural, expressive voice with strong control of intonation and emotional range
  • Access to professional or near-professional recording setup with quality microphone and quiet environment
  • Strong diction and ability to follow scripts naturally while maintaining conversational delivery
  • Availability for a single 4-hour session, with possibility of a second session of similar length

Skills this role asks for

south african english accentvoice delivery and tonalityemotional range and vocal variationscript interpretationmicrophone and audio equipmentpronunciation and dictionrecording consistency

What the interview is likely to probe

  1. 1.Accent authenticity and consistency

    Cloning an inauthentic accent defeats the purpose; the AI learns what it records, so accent must be genuine and steady.

    Expect something like: “By take 12 of a customer greeting, you notice your accent drifting from South African toward neutral English. Why is drift problematic for voice cloning?”

  2. 2.Emotional range and natural interpretation

    A flat monotone clones into a robotic voice; variety trains the model to sound human across contexts.

    Expect something like: “You're assigned to record the same customer service phrase 10 times: neutral, warm and welcoming, concerned and empathetic, then slightly frustrated. How does your vocal delivery change in each version?”

  3. 3.Script precision with natural delivery

    Over-enunciation or reading-aloud style sounds artificial; the goal is conversational naturalness within script boundaries.

    Expect something like: “The script says, 'Your account has been updated with your new billing address.' You could read this stiffly to ensure clarity, or deliver it conversationally. Which approach produces better training data, and why?”

  4. 4.Recording quality and technical setup

    Poor audio (background noise, inconsistent levels, clipped peaks) makes voice cloning worse; you're responsible for your environment.

    Expect something like: “During your fourth hour of recording, you notice traffic noise from a nearby road is seeping into your microphone. You could finish the session and submit files, or pause and try to minimize the noise. What's the right call?”

  5. 5.Multi-take variation and consistency

    The model trains on multiple versions; each take should vary emotionally or in emphasis, but maintain the same accent and voice quality.

    Expect something like: “You're recording the phrase 'I understand your frustration' five times. The first is empathetic, the second matter-of-fact, the third slightly warmer. On takes 4-5 you realize your voice is getting tired and higher-pitched. How do you handle this?”

Exercise you may get

Record or simulate recording three versions of a short customer service script (30 seconds) with different emotional tones, then assess your own consistency of accent, voice quality, and naturalness across takes.

How to prepare

  • Practice reading scripts aloud with natural delivery, varying emotion and emphasis while keeping your accent steady
  • Set up and test your recording space for background noise, microphone positioning, and audio levels
  • Listen to South African voice actors or broadcasters and notice how they maintain natural cadence while delivering professional content

Facts

Pay
$50–100/hr
Commitment
hourly
Work arrangement
remote · Remote
Eligible locations
ZAF
Domain
Language and Audio
Posted
9/24/2026
Open slots
5