OpenTrain is hiring part-time contractor annotators to create dense, multi-tier captions and sub-second trajectory labels for first-person video. C1 English, strong timestamping skills, and availability for a February pilot required; pay is USD 8/hour.
Image & Video Annotation
100% Remote Hourly · $8/hr
$8/hr
Compensation
Worldwide
Eligibility
Intermediate
Experience
Feb 2, 2026
Posted
Open worldwide
Interested in this role?
Create a free OpenTrain account and apply in minutes.
OpenTrain is the hiring and contracting organization for this project. We are the #1 platform for building careers in AI training and data labeling — where contributors join remote projects, develop skills, and work on the human side of building AI.
This role is offered directly by OpenTrain. You will use the OpenTrain platform to complete annotations and pass a short qualification before starting.
Why AI training and this work matters
AI training (data labeling/annotation) is how people teach models to understand the world: by producing and verifying the examples models learn from. This work is remote, flexible, and lets you contribute to state-of-the-art video understanding systems.
This project focuses on egocentric (first-person) video: annotators create high-precision captions and motion trajectories that directly benchmark and improve model capabilities.
The role — overview
Role: Video Annotator / QA (contractor, part-time). You will annotate first-person egocentric video using a 3-tier captioning scheme and produce high-accuracy timestamps and trajectory-level motion descriptions.
Project scope and timing: Pilot in February with target completion for the pilot phase by end of February. Initial volume: 1,000 hours of video with potential to scale up to 10,000 hours based on pilot performance. Quote response requested by January 30, 2026.
Employment type: Contractor, Part-time
Time requirement: 20+ hours/week
Pay: USD 8 per hour
Tool: OpenTrain platform
What you'll do
Annotate dense egocentric videos according to detailed customer guidelines and acceptance criteria (description accuracy, completeness, and timestamp precision). Work is evaluated to a target accuracy of 95%+.
Follow a three-tier captioning scheme and improve pre-labeled segments where provided.
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; these segments may be pre-annotated and require refinement.
Tier 3: Trajectory-level annotations — sub-second, body-part-level motion descriptions (may overlap across limbs), grounded only in visible motion, created from scratch.
Produce precise timestamps (sub-second precision preferred) and atomic, observable action labels (e.g., “grasp cup,” “place lid”).
Requirements and screening
Applicants must confirm C1 English proficiency (comfortable writing precise, natural descriptions) and have prior experience with action-level video annotation and temporal labeling.
You must be able to ramp quickly in the OpenTrain tool and pass a short qualification check before starting the pilot.
C1 English proficiency or higher — strong written comprehension and precision.
Experience with video annotation, action segmentation, and timestamping with tight start/end alignment.
Ability to write verb + object action labels and sub-second trajectory descriptions (left hand/right hand/torso, etc.).
Proven quality performance on similar projects targeting 95%+ accuracy.
Availability to support a February pilot (availability to begin on or around February 3).
How the pilot and scaling work
This posting is for the pilot phase (February). Successful pilot performance may lead to scale-up from 1,000 hours to as much as 10,000 hours of video. Acceptance standards and quality metrics will be enforced for scale.
Annotators will receive full task instructions and guidelines; pre-labeled data will be provided for some Tier 2 work and must be improved to meet acceptance criteria.
Pilot start: February (support needed for pilot in early February).
Pilot target completion: End of February.
Acceptance criteria: description accuracy, completeness, and timestamp precision with target overall accuracy 95%+.
How to apply and what to include
Because this outreach is for planning and pilot selection, we need specific information from applicants. Please respond promptly — quote requested by January 30, 2026.
Provide clear, concise answers to the requested items so we can evaluate feasibility and plan resourcing for the pilot and potential scale.
Confirm C1 English level and relevant video annotation experience.
Provide your indicative hourly rate in USD (we will record USD 8/hr is the listed pay for this posting).
State expected annotation throughput (approximate annotation time per hour of video) and recommended number of resources for a 1,000-hour pilot and for scaling to full production.
State expected ramp-up time, any operational constraints, and earliest available start date (ideally available for pilot in early February).
Join a long-term, remote video-annotation contract to help train vision-language action models by classifying activities and marking objects in everyday household videos using Encord. Fixed-price contract ($58,000 USD) with up to 180 hours/month per annotator and ongoing work for top performers.
Annotate brief driving videos to train autonomous vehicle behavior recognition; $18/hr as a part-time contractor. Must have 1–2 years of labeling experience and a valid U.S. driver's license (IDs may be redacted).
Annotate short soccer video clips by identifying gameplay actions and writing clear 1–2 sentence captions; requires strong soccer knowledge, good English, and attention to detail. Remote, part-time contract work (under 20 hrs/week) at USD 3.50/hour.
