EN-US audio transcription & annotation — large-scale AI data project

We are sourcing en-US transcriptionists for a large-scale audio annotation project running under a major AI data initiative. Work is ongoing and volume is high. About the project You will work in a dedicated web-based annotation tool (Myriad) processing short user-generated audio recordings, with an average duration of approximately 11 seconds per file. The project involves two phases: a pilot phase already underway, and a large production phase targeting 5,000 hours of processed audio. Your tasks Listen to short voice recordings and produce two transcript versions per file: a spoken-form version (lowercase, no punctuation, verbatim) and a written-form version (proper capitalization, punctuation, digits for numbers) Classify each speaker by gender and nativity Determine whether each file should be saved or discarded Apply predefined audio tags (background noise, crosstalk, whispered speech, filled pauses, and other speech/audio conditions) Productivity and compensation Compensation is calculated on a x10 ratio: for every 1 hour of actual audio processed, you are paid for 10 hours of work. For example, 1 hour of audio = 10 billable hours at your agreed rate. Each 11-second audio file requires approximately 110 seconds of work (roughly 1 min 50 sec). Requirements Native or near-native EN-US speaker (C1 proficiency minimum) Located in the United States (location verification required) Strong attention to detail and ability to follow strict annotation guidelines Reliable internet connection and availability for weekly deliveries Ability to commit to a consistent number of hours per week Workflow Training will be provided before project start (planned for the Tuesday before kickoff) QA spot checks will be conducted on your first 10 annotations, with feedback provided Random spot checks will continue throughout the project Work is assigned via a task queue in the Myriad tool — as long as tasks are available, you can continue working Deliveries are organized in weekly batches