Audio Edit Quality Annotator for Speech Data

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We connect people to remote, flexible work that helps train the next generation of AI systems.

As the hiring organization for this role, OpenTrain manages the project, provides instructions and tools, and pays contributors for completed labels.

About AI Training Work

AI training (also called data labeling or annotation) is the human side of building AI: people prepare and review examples that models learn from. This role contributes directly to improving speech models by judging the quality of edited audio.

Many projects are remote, flexible, and accessible — they reward attention to detail and subject-matter familiarity rather than formal university degrees.

• 100% remote work — join from anywhere with a reliable internet connection.
• Flexible contractor work that can fit around other commitments.
• Work directly shapes how speech systems behave in production.

The Role

You will review and label 5 hours of edited speech audio to assess the quality of edits. Each item requires comparing the edited audio to its unedited transcription and waveform, classifying overall edit naturalness and correctness, marking specific issues, and providing segmentation for areas with bad edits.

This is an intermediate-level contractor role for contributors with some experience in audio editing and comfort using audio tools and labeling software.

• Project size: 5 hours of recorded audio to be reviewed and labeled.
• Label types used: Classification and bounding-box style segmentation (time-region marking).
• Labeling tool: Label Studio.

What You'll Do

• Listen to edited audio and review the corresponding unedited transcription and waveform.
• Classify each clip for naturalness and correctness of edits (e.g., seamless, minor artifact, obvious cut, mistranscription).
• Identify and mark specific issues such as abrupt cuts, missing words, misalignments, or artifacts.
• Provide time-based segmentation for regions with bad edits (start/end times or bounding boxes on waveforms).
• Follow detailed instructions and quality guidelines to ensure consistent labels across the dataset.

Requirements

You must have hands-on experience with audio editing and be comfortable working in audio-editing software. Strong attention to detail and the ability to follow detailed instructions are essential.

• Some experience with audio editing (required).
• Basic proficiency using audio editing software (e.g., Audacity, Adobe Audition, Reaper).
• Experience with labeling tools such as Label Studio is a plus.
• Reliable internet connection and ability to meet project deadlines.
• Intermediate experience level — able to judge subtle edit artifacts and follow complex guidelines.

Compensation & Work Details

This is contract work paid per label. Rate: $0.1275 USD per label. You will be paid as a contractor according to OpenTrain's payment schedule.

The project is open worldwide. There is no fixed schedule provided in the brief; you'll complete items according to the project's workflow and deadlines.

• Payment type: Pay-per-label at $0.1275 USD per label.
• Employment type: Contractor (remote, worldwide).
• Tool required: Label Studio access; use of your preferred audio editor for listening/segmentation.

How To Apply

Create an OpenTrain account (free), complete your profile, and apply to this project. In your application, highlight your audio-editing experience and any familiarity with labeling tools.

If selected, you will receive project instructions, access to the labeling interface, and a dataset sample to begin work.

• Include brief examples of prior audio-editing work or labeling experience when you apply.
• Be ready to follow a quality checklist and sample-based training before labeling full items.