Annotator

Role Overview:

We are seeking detail-oriented listeners with hands-on experience in audio captioning, transcription, audio production, or voice and linguistic analysis. This role supports the development and benchmarking of audio-capable Large Language Models (LLMs) by producing precise, evidence-based descriptions of speech, non-speech sounds, and mixed audio scenes. You will help ensure AI-generated captions reflect what is actually audible, not what a video, topic, or draft wording implies.

Required skills and qualifications:

• Experience in audio captioning (ideal), transcription, or audio annotation.
• Background in linguistics, voice analysis/coaching, or audio production is preferred.
• Native-level written English with the vocabulary to describe voices and sounds precisely and concisely.
• Ability to distinguish English accents and recognize other languages.
• Strong listening acuity: able to hear quiet, brief, overlapping, and changing details and separate individual sources.
• Describes sound in everyday words: what it sounds like, how loud it is relative to other sounds, and when it starts, stops, or changes.
• Sustained focus and consistency across long batches.

Task Overview:

• Listen critically to audio clips and write clear, concise, natural English descriptions of voices, sounds, and scene changes, without adding unsupported assumptions.
• Transcribe English speech faithfully, including repetitions, fillers, and names, and flag lines spoken in another language.
• Track who speaks when, attributing each line to the correct speaker through overlaps, interruptions, and group speech, and recognize when a new voice enters.
• Describe how each voice sounds and how its delivery changes (pitch, voice quality, loudness, pace, emotion, intonation), judging delivery from the sound rather than the meaning of the words.
• Identify, count, and time non-speech vocal sounds (laughs, coughs, sneezes, sobs, gasps, breaths) and other sounds by texture, attack, distance, and relative level, distinguishing discrete events from continuing ambience.
• Judge which sounds stand out or mask others, and whether a voice or sound belongs to the recorded scene or was added over it (narration, playback, an added music bed).
• Place onsets, offsets, speaker changes, overlaps, and gradual transitions accurately.
• Review machine-generated drafts critically, catching invented, missing, mistimed, or misattributed content.
• Apply consistent judgment throughout each clip and across each batch.