Audio Segmentation and Emotion Transcription

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. We help people start and grow careers teaching AI by connecting contributors with projects, providing tools and clear instructions, and supporting flexible, remote work.

About AI training and this work

AI training (also called data labeling or annotation) is the human side of building intelligent systems: people prepare and evaluate examples the models learn from. This project places you on the front lines of that work by helping models understand emotion and expressive speech.

This kind of work is often remote and flexible, making it a strong fit for part-time contributors, students, or anyone who wants to earn while working from home. Many projects need no prior experience beyond strong language skills and attention to detail.

The Role

You will listen to 1–2 minute English audio recordings, identify 10–12 second clips where speakers convey strong emotional expression, and assess emotional valence (positive/negative) and arousal (intensity) using provided Likert scales. For each chosen clip you will submit two transcriptions: a detailed verbatim transcription that captures verbal content and emotional delivery, and a concise summary.

This is entry-level contractor work, under 20 hours/week, paid per hour at USD 8.00. The project values careful judgment: instructions are intentionally broad to allow you flexibility in interpreting emotional cues, so strong listening and writing skills are essential.

What You'll Do

• Listen to short (1–2 minute) audio files and locate 10–12 second segments with clear emotional expression.
• Rate each selected clip’s emotional valence (positive/negative) and arousal (intensity) using provided Likert scales.
• Create a detailed verbatim transcription that captures words, tone, pauses, pitch changes, and other emotional cues.
• Write a brief summary transcription that captures the main content and emotional takeaway.
• Avoid selecting neutral or low-arousal content; focus on clips where emotion is distinctly conveyed.
• Work independently and follow evolving or experimental instructions as provided.

Requirements

We need careful listeners with excellent English writing and transcription skills who can identify subtle emotional cues in speech and make sound judgment calls. This role is suitable for people with prior transcription or labeling experience but is open to entry-level applicants who meet the language and attention-to-detail requirements.

• Excellent written English and strong transcription skills.
• Ability to detect and describe emotional valence and arousal using Likert scales.
• Strong attention to detail and sensitivity to tone, pauses, and intensity in speech.
• Experience with audio transcription and/or annotation is preferred but not required.
• Comfortable working with experimental instructions that may change and making independent judgments.
• Must be able to create both long (detailed) and short (summary) transcriptions.

Test & Application Process

Candidates will complete a short test during the interview. You must submit your test transcription before the interview is concluded.

Test instructions and guidelines are below; follow them closely and submit both the detailed transcription and the brief summary when asked.

• Test task: Transcribe the first 10 seconds of this YouTube clip: https://www.youtube.com/watch?v=8SxE_NfUX6w
• Long transcription guidelines: transcribe verbatim, capture emotional cues (tone, pauses, pitch changes), and optionally describe the speaker’s apparent voice characteristics.
• Do not include background sounds or music—focus on spoken words and emotional delivery.
• Submit your transcription in the chat before the interview ends; the interviewer will not close the interview until they receive it.

Compensation, Schedule & How It Works

This is a remote, contractor, part-time role. Expect less than 20 hours per week with flexible scheduling. You will be paid hourly at USD 8.00 per hour.

We provide the annotation interface and Likert scales for rating valence and arousal; you will access audio files, submit selected segments, and enter transcriptions through the provided tools.

• Employment type: Contractor, Part-time.
• Pay: USD 8.00 per hour (PAY_PER_HOUR).
• Time commitment: Less than 20 hours/week, flexible scheduling.
• Data type: Audio; Label types: Segmentation and Text Generation.
• Worldwide applicants accepted; specific language requirement: excellent English writing and transcription.