Scene Cut Annotation for YouTube Videos

About OpenTrain

OpenTrain is the #1 platform for finding and building careers in AI training and data labeling. Creating an OpenTrain account is free and helps vendors and annotators discover projects, build a profile, and apply quickly.

We connect experienced annotation teams with projects across industries so your work directly shapes how modern AI systems behave.

About AI Training Work

AI training (also called data labeling or annotation) is the human side of building AI. Annotators prepare and validate the examples modern models learn from — in this case, temporal segmentation of video to improve content analysis, summarization, and automated editing.

This kind of work is often remote and flexible, accessible to teams with strong attention to detail and video-editing experience, and puts contributors on the cutting edge of media & entertainment tooling.

The Role

We are hiring a qualified data-labeling vendor or team to perform scene cut and transition annotations on a large batch of YouTube videos. The project focuses on temporal segmentation: marking the exact start and end timestamps for each scene boundary.

Key facts: roughly 1,000 videos (average 5–10 minutes each), preferred team size 10–15 annotators, project duration 4–6 weeks, annotation tool uLabel (access provided), and required final deliverable in JSON or CSV with fields video_id, scene_id, start_time, end_time.

• Data type: Video; labelling type: Segmentation (temporal).
• Annotation tool: uLabel (project will provide access).
• Expected output: JSON/CSV with video_id, scene_id, start_time, end_time.

What You'll Do

Follow the supplied annotation guidelines to identify scene cuts and transitions and record precise timestamps. Perform internal quality checks and revisions before delivering batches to our team.

• Annotate scene boundaries (hard cuts and transitions) and mark start/end timestamps for each scene.
• Use uLabel for all annotations; adhere to guideline examples and edge-case rules.
• Run internal QA on every batch to ensure consistency before submission.
• Deliver final files in the agreed JSON or CSV schema (video_id, scene_id, start_time, end_time).
• Respond to weekly quality audits and address any requested revisions within agreed timelines.

Requirements

We expect teams with prior experience in video annotation or temporal segmentation, the ability to meet high accuracy standards, and capacity to deliver on the stated timeline and volume.

• Proven prior experience with video annotation or temporal segmentation tasks (must be stated in proposal).
• Intermediate experience level: ability to follow detailed guidelines and apply consistent judgments.
• Capacity to staff a team (preferred 10–15 annotators) and process ~1,000 videos in a 4–6 week window.
• Maintain ≥95% accuracy on spot checks and cooperate with weekly internal QA audits.
• Contractor engagement; work is remote and may be performed from anywhere (worldwide).
• Compensation structure: PAY_PER_HOUR at USD 4/hour (as stated in project details).

Proposal & Selection Criteria

Submit a proposal that demonstrates relevant experience, team capacity, quality procedures, and pricing. We will evaluate proposals based on experience, demonstrated accuracy controls, delivery speed, and cost.

• Prior experience with video temporal segmentation and examples or references.
• Estimated team size, daily/weekly throughput, and total delivery timeline.
• Costing: specify cost per annotated video or per hour of video annotated (per the RFP requirement).
• Description of your internal QA process, accuracy monitoring, and how you ensure ≥95% spot-check accuracy.
• Confirmation you can deliver output in the required JSON/CSV schema and use uLabel (access provided if selected).

How Work Is Organized & Next Steps

Selected vendors will receive access to the annotation environment and the project guideline pack. Work will be organized in batches with weekly quality checks from our internal QA team.

Provide a clear proposal to be considered; details on onboarding, exact delivery milestones, and batch schedules will be shared with chosen teams.

• Project duration: 4–6 weeks with batch-based deliveries and weekly QA audits.
• Volume: ~1,000 videos (avg 5–10 minutes each); preferred team size 10–15 annotators.
• Accuracy target: ≥95% on spot checks; vendors must perform internal QC before submission.
• If selected, you will be given access to uLabel and the annotation guidelines to begin onboarding.