Turing is staffing a shot boundary detection project: annotators watch video clips frame-by-frame, mark where each shot starts and ends, label the transition type, and group shots into scene buckets with genre categories. It suits people with editing or production experience who can hold a labeling standard steady across hundreds of clips.

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

What the work actually is

You receive video clips and step through them frame-by-frame in an annotation tool. For every shot you create a "Shot Card" recording the exact start and end frames, then label the transition into and out of it — hard cut, fade to/from black, dissolve, wipe, plus whatever subtypes the guidelines define. Once the shot list is complete you group shots into scene buckets based on narrative or event continuity, and tag both clip and bucket with a content category such as TV-Streaming/Drama, Education/Documentary, Gaming/Sci-Fi, or Podcast/Lifestyle.

The hard part is not the tooling, it is the boundary calls. A whip pan, a fast zoom, a match cut on motion, a hidden cut behind a foreground wipe, a crossfade that lasts four frames, a hard light change that reads like a cut but is not — these are where annotator agreement collapses and where the training signal either gets built or poisoned. Gaming footage adds its own problems: camera respawns, cutscene-to-gameplay handoffs, UI overlays. You are expected to make a defensible call, apply it consistently, and write a short note flagging genuinely ambiguous cases for review rather than silently guessing.

What the screen looks for

Turing asks for 1+ year in video/film editing, content creation, or production, and the screening reflects that: expect questions about transition vocabulary, frame-accurate boundary reasoning, and how you behave when the guideline does not cover the clip in front of you. Film studies, cinematography, or prior annotation experience are listed as nice-to-haves, not gates. Shortlisted candidates complete an assessment before starting — in practice this means annotating sample clips against a rubric, so your calibration against a written spec matters more than your taste.

Logistics

Remote, flexible hours, with quality and throughput targets you are expected to hit independently
Compensation is described as competitive but no band is published; treat any figure you hear as observed for a specific cohort, not guaranteed
Structured guidelines are supplied and updated; expect to re-read them and to have earlier work spot-checked
Volume work — the role rewards people who can sustain frame-level attention over long sessions, not just on a good clip

Video Annotator

What the work actually is

What the screen looks for

Logistics

Interview prep — 10 questions