Egocentric Video Annotation - C1 English Required

About OpenTrain

OpenTrain aggregates data-labeling and AI-training jobs from many companies into one job board so contributors can find and apply to projects in one place. Creating an OpenTrain account is free and applications take only a few minutes.

We list projects like this one so skilled annotators, reviewers, and QA specialists can discover remote, flexible work that directly shapes how AI systems perform.

About AI training and this kind of work

AI training (also called data labeling or human feedback work) is the human effort that teaches models to understand actions, language, and visual content. Tasks include annotating video, transcribing audio, writing and rating model outputs, and reviewing model behavior.

This project involves detailed, quality-focused video annotation: it requires careful observation, precise timestamps, and clear natural-language descriptions that models will learn from.

The role

You will annotate first-person (egocentric) videos on SuperAnnotate following detailed customer guidelines. Work is contractor/part-time and requires strong written English (C1) and prior video annotation experience.

This posting is for a pilot in February with potential scale-up from 1,000 hours to up to 10,000 hours of video based on pilot performance.

• Project name: Human Generated Egocentric Annotations (Action and Trajectory Level Annotations).
• Data type: Video (egocentric) with pre-labeled action segments to improve and new trajectory labels to create.
• Label types: Text generation (captions), action recognition (temporal segments), and tracking/trajectory-level motion descriptions.
• Annotation tool: SuperAnnotate (or equivalent familiarity required).

What you'll do

Follow customer-provided guidelines to produce three-tier captions and trajectory descriptions with high precision and completeness.

• Tier 1: One high-level video summary (1–2 sentences, no timestamps).
• Tier 2: Action-level segments with start/end timestamps and verb+object labels (pre-annotated segments will be improved).
• Tier 3: Trajectory/body-part-level motion descriptions created from scratch: sub-second segments, possible overlaps, grounded only in visible motion (no intent).
• Deliver work meeting acceptance criteria on description accuracy, description completeness, and timestamp precision (target 95%+).
• Participate in a short qualification check before starting and follow review/QA workflows as required.

Requirements and screening

You must confirm C1 English proficiency or higher and demonstrate prior experience with action-level video annotation and timestamping with tight start/end alignment.

• Confirm C1 English proficiency (comfortable writing precise, natural descriptions).
• Prior experience with video annotation (action segmentation / temporal labels) and timestamping with sub-second precision preferred.
• Ability to write atomic observable action labels in verb + object format (e.g., "grasp cup").
• Ability to generate trajectory/body-part-level motion descriptions (left hand/right hand/torso) with <1s segments and overlaps.
• Familiarity with SuperAnnotate or equivalent annotation tools and ability to ramp quickly.
• Proven quality performance on similar projects targeting 95%+ accuracy and low rework.
• Availability to support the pilot starting February 3 and to scale if selected.

Project details, timeline, and acceptance criteria

Key dates and targets are strict due to this RFQ and pilot schedule. Please review and confirm you can meet these timelines before applying.

• Target completion date (pilot baseline): end of February.
• Pilot start: availability to support a pilot in February (pilot resources expected to be active on February 3).
• Quote needed by: January 30, 2026 (this outreach is for estimation/planning only; no commitment is required at RFQ stage).
• Volume: initial pilot ~1,000 hours of video with potential to scale up to ~10,000 hours.
• Acceptance criteria: description accuracy, description completeness, and timestamp precision (project target: 95%+ accuracy).
• Annotation tool: SuperAnnotate platform (customer-provided guidelines and assets).

Compensation, schedule, and contract

This is a remote contractor, part-time role. The listed base pay is per hour; applicants should confirm their availability and expected hourly throughput.

• Payment type: Pay per hour.
• Listed hourly rate: USD 8.00 per hour (PAY_PER_HOUR).
• Time requirement: 20+ hours per week during the engagement or as agreed for pilot staffing.
• Employment types: Contractor, Part-time; global applicants accepted (any location).

How to apply and what to include

Because this RFQ is for estimation and pilot planning, provide concrete inputs so the requester can evaluate feasibility. Submit answers and supporting details along with any portfolio or work samples that demonstrate similar video annotation experience.

• Confirm C1 English proficiency and describe how you validate it (certificates, past work samples, or a short writing sample).
• List relevant experience with action segmentation, tight timestamping, trajectory/body-part annotations, and SuperAnnotate (or equivalent).
• Provide your indicative hourly rate in USD (if different from the listed rate), your recommended skill level for the role, and the number of resources you'd propose for the pilot.
• Estimate resources needed if the project scales to full production (up to 10,000 hours) and give any assumptions used to compute that estimate.
• Share any early assumptions on annotation time per hour of video (e.g., average minutes of annotation per 1 video hour) and expected ramp-up time to full throughput.
• Note operational constraints, preferred shift/hours, and readiness to begin the pilot around February 3.
• Attach or link to relevant reference material or past project examples and be ready to pass a short qualification check before starting.