Detecto
Video │ file-based

Per-scene AI video detection, on the file you uploaded.

Drop in a video file. Detecto returns a generation likelihood, a per-scene timeline, audio-vs-frame divergence (when sound is present), and a reviewer-friendly scene grid. v1 is file-based; no URL ingestion or live-stream callbacks.

Try a scanHow calibration works
Try a scan

Drop a file here, or pick one

This is a quick demo, files are not saved. Sign in to keep your reports.

5 credits per 10 seconds

No account needed. Anonymous scans are rate-limited per IP.

Built for the people who review other people’s content

  • Editors at independent magazines
  • University writing centers
  • Newsroom verification desks
  • PR review teams
  • Content QA leads at agencies
File-based
Accepted inputs

Video files, ≤ 50 MB.

MP4, WebM, MOV. Designed for short clips. Slice longer assets into the segments your decision turns on.

Citizen-submitted clips
Phone-shot witness footage, on-the-scene short clips.
Editorial b-roll
Vendor-supplied b-roll and stock-like clips for editorial use.
Verification escalations
Video flagged by a first-pass reviewer for a second look on a verification desk.
PR-supplied executive cuts
Executive interviews, brand-supplied footage, agency-produced cutdowns.
What we measure

Signals exposed in every video report.

Aggregate
Generation likelihood
Single likelihood across the file. Read alongside the per-scene timeline, never on its own for mixed-source video.
Per-scene
Per-scene timeline
Likelihood plotted per detected scene, surfaces synthesized cutaways inside otherwise authentic footage.
Audio
Audio/video divergence
Higher values indicate de-synced or synthesized audio overlaid on real frames, or vice-versa.
Surface
Frame-level artifacts
Cumulative artifact count from per-frame analysis (skin, hands, lighting, micro-textures).
Metadata
Container integrity
Container-level metadata, codec inconsistency, and re-encode trace flags.
Calibration
Confidence variance
Variance across resampled detection passes; flagged when the per-scene timeline is volatile.

Honest limitations on video detection

A detection score is probabilistic, not proof. Treat the report as evidence to discuss, not a verdict to publish.

  • v1 is file-based. No URL ingestion or live-stream callbacks.
  • Designed for short clips; long-form video is out of scope until the async media path is fully provisioned.
  • Heavy compression (re-encoded social re-shares) erodes per-scene signal.
  • Mixed-source video (real footage with synthesized cutaways) requires the per-scene timeline; the aggregate score is the wrong number to read alone.
  • Audio-vs-frame divergence depends on the audio track being present and clean.
For verification desks

The per-scene timeline lands before the article does.

Your reports save privately to your dashboard, shared only when you choose.

Get startedFor publishers