Video Data Annotator

$7/hrrate as listed

Label objects, actions, and events frame-by-frame in video clips used to train and evaluate computer vision and multimodal models. Work is remote, task-based, and paid at an observed rate of around $7/hour through micro1's contributor pipeline.

Summary, eligibility guidance and screening preparation are prepared by Sidequest from the platform's posting. Check the source listing before applying.

What the work involves

Video annotation is repetitive, detail-heavy labeling work performed in a browser-based tool. Depending on the project you are routed to, a shift might involve drawing and adjusting bounding boxes across frames, tracking a person or vehicle through occlusion, marking the exact frame where an action starts and stops, segmenting objects at the pixel level, or writing short natural-language descriptions of what happens in a clip. Most projects come with a written guideline document — often 10 to 40 pages — and your output is judged against that document rather than against your own intuition about the video.

Frame-level and clip-level labeling: boxes, polygons, keypoints, temporal segments
Object tracking with consistent IDs across long sequences
Action and event tagging, including boundary precision to the frame
Flagging unusable clips: motion blur, bad lighting, ambiguous content, guideline gaps
Reworking batches that fail QA and responding to reviewer comments

What the platform screens for

micro1 runs an AI-led interview before routing anyone to paid work. For annotation roles the screen is less about credentials and more about reliability and instruction-following: can you read a spec carefully, apply an edge-case rule consistently on item 400 the same way you did on item 4, and explain your reasoning when two labels both look defensible. Expect questions about tools you have actually used (CVAT, Label Studio, SuperAnnotate, V7, Encord, or an in-house equivalent), your throughput on past projects, and how you handled a disagreement with a reviewer. Clear spoken and written English matters, since many projects include free-text captions and reviewer threads. A stable internet connection and a machine that can scrub HD video without stuttering are practical prerequisites, not nice-to-haves.

Logistics and honest expectations

This is remote and largely asynchronous. Task batches open and close, so availability windows matter more than a fixed schedule — contributors who can commit 15 to 30 hours a week and respond within a day to QA feedback tend to stay in the queue. Pay is stated as observed at roughly $7/hour and is not guaranteed; rates vary by project, geography, and whether a task is billed hourly or per unit. Volume is genuinely inconsistent: a project can run for weeks and then end abruptly when the client's data need is met. Treat this as a supplementary income stream with a path to higher-paid QA and lead-reviewer work rather than as a stable full-time role.

What they look for

Video Data Annotation
Data annotation
Attention to detail
AI Review

Before you apply

Would you pass this screen?

Drawn from what this listing actually requires. Answered in your browser — nothing is sent to us, and we do not store your answers.

Have you used at least one professional video annotation tool (such as CVAT, Label Studio, SuperAnnotate, V7, Encord, or equivalent) to label video data in a real project?
Annotation platform experience · required
Do you have a stable broadband internet connection (minimum 10 Mbps download, 5 Mbps upload) and have you confirmed that your machine can play and scrub HD video without stuttering?
Stable internet connection · required
Can you commit to 15 or more hours per week to annotation work and respond to QA feedback within 24 hours?
Availability commitment · required
Are you able to read and write clear English, including understanding detailed technical guidelines and responding to reviewer comments in writing?
English fluency · preferred

0 of 4 answered

Included with this listing

Interview prep — 11 questions

Matched to what this platform's screening actually measures. Answer each out loud in 60–90 seconds: claim → concrete example → implication.

What it tests: Whether you have hands-on experience with a real annotation platform and can apply judgment consistently when guidelines are ambiguous—not whether you memorized tool buttons.
A strong answer: Name a specific tool (CVAT, Label Studio, V7, Encord, SuperAnnotate, or equivalent) and a concrete labeling task (bounding boxes, tracking, segmentation, temporal boundaries). Describe one moment where the spec left room for interpretation, explain how you resolved it (by returning to the core intent of the guideline or flagging it), and show that you applied that same logic consistently afterward.
Where people lose it: Saying you've used a tool without being able to describe a real project, or claiming you just followed your gut when guidelines were unclear—annotation is spec-driven, not intuition-driven, and reviewers will notice if your frame 400 contradicts frame 4.