Audio Captioning Expert at Mercor

Audio Captioning Expert | $30–35/hr | Remote Worldwide

Mercor is partnering with a leading AI research lab to recruit freelance audio annotation specialists for Project Apollo, a large-scale media annotation initiative focused on improving AI understanding of audio in video content. This high-volume, fast-paced opportunity is ideal for detail-oriented professionals who excel at structured listening, transcription, and analytical audio labeling.

Role Overview

As an Audio Captioning Expert, you'll watch short-form videos (0–3 minutes) and produce high-precision audio annotations following strict formatting and quality guidelines. Your work will center on careful human analysis of audio signals, including speech, music, sound effects, and emotional tone within video content.

Key Responsibilities

• Annotate video content with primary focus on audio analysis: speech segmentation and delivery style, music identification and detailed description, sound effects, and key non-speech audio
• Evaluate emotional arc derived from audio and visual cues
• Apply rubric-based criteria to ensure accurate and consistent audio content evaluation
• Ensure all annotations are grounded strictly in audible or visible content, chronologically precise, selective, concise, and free of speculation
• Reject unusable samples according to project rules (e.g., no audio or non-English speech)
• Collaborate asynchronously with research team to maintain and improve annotation quality standards

Ideal Qualifications

• Strong ability to analyze and describe audio content with precision
• Excellent written communication skills and exceptional attention to detail
• Experience in audio annotation, transcription, media labeling, content moderation, or qualitative analysis
• Ability to quickly process and assess diverse short-form video content
• Comfort working within strict formatting and quality guidelines
• Reliable internet access and distraction-free work environment

Work Standards

• All annotations must reflect original human judgment
• Use of LLMs (ChatGPT, Claude, Gemini) for writing, rewriting, or evaluating annotations is strictly prohibited
• Accuracy, selectivity, and guideline adherence are non-negotiable

Commitment Details

• Expected commitment: 40 hours per week
• Ongoing project with potential for long-term engagement based on performance
• Hourly rate: $30–35/hour