What the work involves

You record scripted German audio across several registers — conversational support dialogue, instructional walkthroughs, narrative passages, and short corporate-tone lines. Sessions typically involve batches of prompts with multiple takes: the same line delivered calmly, then energetically, then with different emphasis placement, so the model learns prosodic variation rather than a single flat read. Consistency matters more than performance flair here; your voice, accent, and mic distance need to match across sessions that may be weeks apart, because inconsistency shows up as artifacting in the synthesized output.

Delivery guidelines are detailed and non-negotiable: specified sample rate and file format, a noise floor below a stated threshold, no room reverb, pop filter in place, and clean file naming. Expect to re-record takes flagged by QA for breath noise, sibilance, mouth clicks, or drift in pace. Recordings are used exclusively for the client's internal CX AI Agent and are not sold, licensed, or folded into other datasets — but the work does involve voice cloning, and you should only apply if you are fully comfortable with a synthetic version of your voice being deployed in that context.

What the platform screens for

Mercor's screening is AI-led and leans on verifiable specifics. Expect to name actual projects — dubbing credits, audiobook titles, IVR or TTS datasets you've contributed to — and to describe your signal chain by make and model rather than in general terms. Interviewers probe whether you can articulate why a take failed, not just that it did, and whether you understand what makes audio usable as training data versus merely pleasant to listen to. Germany-based residency and native German fluency are hard requirements and will be checked. Explicit, informed consent to voice cloning is confirmed on the record.

Logistics

  • Fully remote, recorded from your own space; asynchronous within weekly batch deadlines
  • 5–10 hours per week, sustained over the project duration
  • Observed rate: $50/hr — a band reported for this listing, not a guaranteed offer
  • Familiarity with Audacity, Reaper, or Adobe Audition is preferred; light self-editing may be requested
  • Ability to switch between vocal styles on request is a meaningful advantage