Banned user returns with a profile video
Different email, different username. Face recognition identifies the match from the video frames.
Match · banned user · 96%
Face recognition in video
Match faces in video content against known identities. Lasso extracts frames, encodes faces, and compares them to your database of banned or verified users.
Removed video re-uploaded by a different account
Frame-level hash matching detects the duplicate even after re-encoding, cropping, or trimming. Blocked automatically.
Hash match · removed contentVideo selfie verification on a dating platform
Frames matched against profile photos. Identity confirmed across multiple angles in a single recording.
Verified · matches profile · 98%Every face checked against the verified performer database
Unverified individuals flagged before the video goes live. Compliance with content source verification requirements.
Unverified face · flagged
Video face recognition: four layers from frame to identity
1ML detection
Frame extraction and encoding.
Faces detected in each frame. Face encoding generated per face for identity matching. Image hash generated per frame for duplicate detection.
See Lasso in action
Face encoded — frame 0096
Frame Extraction
Encoding complete
Faces detected
1 face · 24 frames sampledFace
Face encoding
Generated per face, per frameEncode
Frame hash
Generated per frameHash
Encoding confidence97%
2Custom rules
Face encodings from video frames compared against your banned user database.
Frame hashes compared against your disallow list. Match thresholds and actions configured per platform: auto-reject, flag for review, or escalate.
See Lasso in actionLayer 1 Output
1Encoding vs banned users: match 96%
2Frame hashes vs disallow list: clear
3Verified user database: no match
Custom Rules
BlockBanned user database
Face match above 90% threshold
Frame hash on disallow list
Match between 70–90%: flag for review
Match on watchlist: escalate
Auto-rejected
Account flagged · video never goes live
3AI Moderator
Context-aware matching across video frames.
Low-confidence match on one frame may become high-confidence when the same face appears from multiple angles throughout the video. AI Moderator evaluates the full recording.
See Lasso in action
Second face — no match
AI Moderator
Analyzing
Frame analysis
2 faces across 162 frames · upload, not live
Database match
Face 1 matches verified performer · 98%
Anomaly detected
Face 2: no database match · one frame alone, 58%
Cross-frame aggregation
Same face from multiple angles — aggregated 74%, still unverified
Escalated
Unverified face · below auto-action threshold
4Human review
Human review with full video context.
Matched frames shown alongside database photos. Confidence scores per frame. Account metadata. Every human decision improves future matching accuracy.
See Lasso in actionQueued for review
AI Moderator reasoning
A second face in the recording matches no verified performer. Aggregated confidence across every angle stays below threshold — needs a human check.
Human Review1 of 3
Match data
Matched frames shown beside database photos · Face 2: 0:08–2:41
AI assessment
Unverified individual in frame — face 1 is a verified performer, face 2 matches no approved identity
AI confidence
Identity match74%
Below auto-action threshold (90%)
This decision will train the AI for future similar content
Five capabilities included in Lasso's video face recognition.
Match faces from video frames against a database of known identities. Lasso extracts the frames, encodes each face, and compares it to your banned or verified user database, holding up across different angles, lighting, and appearances.
Video uploadframes
front0:02
three-quarter0:09
tilted · low light0:14
Known identitiesdatabase
User #4821banned · Mar 2026banned
User #1198verified · creatorverified
User #3377watchlistwatch
Encoding match
96%frame 0:09 ↔ #4821
Re-offender detection extended to video. Banned user returns with a profile video instead of a photo. Face recognition identifies the match from video frames. Account flagged before any content goes live.
Video performer verification for adult platforms. Every face in every frame compared to approved identities. Content with unverified performers blocked or flagged. Compliance audit trail included.
A single image shows one angle, while video gives the AI Moderator dozens. It uses every frame to build a more confident identity match and resolves edge cases with account context.
Frame-level hash matching for video. Previously removed content re-uploaded under a new account. Hash matching on extracted frames identifies the near-duplicate. Two layers of re-offender detection: faces and content.
Video uploadframes
front0:02
three-quarter0:09
tilted · low light0:14
Known identitiesdatabase
User #4821banned · Mar 2026banned
User #1198verified · creatorverified
User #3377watchlistwatch
Encoding match
96%frame 0:09 ↔ #4821
Lasso: next-gen AI content moderation
99.9%
On autopilot, and getting smarter every day.
Three layers of AI handle the volume, your rules, and the grey areas. Your team only sees what truly needs them, and every decision they make improves the system.
Keep users safe without driving them away.
Customizable moderation that lets you find the right balance between safety and user experience. So you protect your community without suppressing the culture that makes it worth joining.
Complexity removed from content moderation.
One API. Clear dashboards. A moderation pipeline built around one-click actions and the right context, right where you need it.
★★★★★
4.9
Highest rated in content moderation on G2.
Every platform has its returning offenders. Lasso recognizes them.
Adult entertainment
Performer verification on every frame, not just the upload form.
- Content with unverified performers appearing in uploaded video.
- Previously removed videos re-uploaded after re-encoding.
- Performers using unverified accounts to bypass content source verification.
Dating
Catch banned users hiding behind new video profiles.
- Users banned for harassment creating new accounts with video profiles to avoid photo-based detection.
- Stolen video introductions used for catfishing.
- Previously flagged profile videos re-appearing on new accounts.
Social platforms
Re-offender detection at UGC video scale.
- Creators banned for policy violations returning with video content on new accounts.
- Same videos re-uploaded after minor edits.
- Repeat offenders evading account-level bans through new registrations.
FAQs
Video face recognition, answered
Every match comes with a confidence score per frame, so you set the threshold that triggers an action. Because a single recording supplies dozens of frames from different angles, the system aggregates them into a more confident identity match than any one photo could give. Anything below your threshold routes to human review with the frames and database photo attached.
Usernames and IPs are easy to change: a banned user simply registers a new account. A face is far harder to swap. Lasso matches the face in the video frames against your banned-user database, so returning offenders are caught even when the email, username, and device are all new.
Yes. Lasso generates an image hash per frame and compares it against previously removed content, so a near-duplicate is flagged even after re-encoding, cropping, or trimming. This works alongside face recognition, giving you two layers of re-offender detection: the faces and the content itself.
Lasso compares face encodings rather than storing raw video, and every match and review decision is written to an audit log. That trail supports DSA reporting and performer-verification requirements, and the platform is GDPR-compliant. You control which databases faces are matched against and how long records are kept.
Video face recognition identifies people in video content by matching their faces against a known database. Frames are extracted from the video, faces detected and encoded per frame, then compared to your database of banned, verified, or flagged users.
Face detection counts faces and reads attributes like age and gender per frame. Face recognition asks who the person is, matching each face against your database. The two run on the same pipeline, applied frame by frame.
Yes. It holds up across a different angle, lighting, or hairstyle because multiple frames capture the same face from several perspectives, which raises match confidence.
Your rules determine the response: auto-reject, flag for review, or log, with different actions at different confidence levels. Moderators reviewing a flagged match see the extracted video frames alongside the database photo, confidence scores per frame, and account history.
Performer verification on uploaded video for DSA compliance. Every face in every frame compared to approved identities. Unverified performers flagged. Audit trail tracks every match and review decision.
Know exactly who's in your video
Lasso matches faces in video against your banned and verified user databases, catching returning offenders that a username check misses. One pipeline, one API.
Book a demo
Lasso Moderation
© 2026. All rights reserved.
