Vocal Audio Specialist - Freelance AI Trainer Project

Invisible Agency

6d•$6 - $65•Remote

About The Position

We are sourcing independent Vocal Audio Specialists to provide their expertise for an AI audio model training project. As AI models increasingly generate and interpret human musicality and vocal nuances, their accuracy relies entirely on robust, high-fidelity training data. The objective of this project is to autonomously record clean, professional-grade audio detailing human vocal pitch changes by performing consistent, highly controlled humming exercises.

Requirements

Demonstrable musical literacy and performance history (e.g., choir, band, a cappella, live gigs, audio engineering, or music production) with a calm, controlled voice and a highly trained ear for pitch stability.
Hands-on experience in Voice Over (VO) work, studio recording, or prior AI audio training projects.
Strong technical recording competence, including a deep understanding of gain staging, microphone technique, and how to operate recording equipment effectively.
A meticulous, detail-oriented approach to following instructions and maintaining rhythmic and tonal consistency across multiple recorded takes.
Must possess a controlled, quiet home recording environment capable of capturing clean audio free from background noise, clipping, or room echo.
Supply a secure computer, professional recording equipment, and high-speed internet.

Responsibilities

Recording clean, pitch-stable humming takes with dynamic control (e.g., humming softly without disappearing or louder without distorting), while strictly following tonal prompts (e.g., "hum a major 3rd," "stay in key").
Demonstrating a solid "internal metronome" to follow beats and executing highly repeatable tasks, such as humming the exact same melody multiple times with minimal variation.
Applying professional recording competence—including proper mic proximity, gain/input controls, and gain staging—to ensure distortion-free, high-fidelity audio capture within a controlled environment.
Accurately executing highly specific vocal directions (e.g., "flat tone, no vibrato") to generate clean, reliable, and standardized data for model training.