Two people in a single-person profile video
Face count averaged across frames. Video rejected, user prompted to re-record.
Face count: 2 · rule allows 1
Face detection for video content moderation
Detect faces across video frames. Lasso returns face count, age range, and gender per frame with confidence scores. Your rules and context-aware AI handle enforcement automatically.
Underage face in an adult-platform upload
Age estimated per frame. Content blocked automatically, uploader flagged.
Age 15–17 · 91% confidenceFaces detected live as the stream plays
Counted and age-estimated live, while the stream plays.
Live · 1 face · age 20sFace filter dropping detection confidence
Low confidence flagged across frames. User asked to re-record without filters.
Low confidence · filter detected
Video face detection: four layers from frame to decision
1Frame extraction & ML
Frames extracted from the video, face detection run on each.
Returns face count, age range, gender, and confidence scores per frame. The same ML pipeline that powers Lasso's image face detection, processed at scale.
See Lasso in action
2 faces — frame 0142
Face Detection
Detection complete
Faces detected
2 faces in frameCOUNT
Age estimate
Face 1: 24–30 · Face 2: 22–28AGE
Gender
Face 1: male · Face 2: femaleGENDER
Detection confidence97%
2Custom rules
Your rules enforced across every frame.
Set face count limits, age minimums, and visibility requirements per content type from the dashboard. No code required.
See Lasso in actionLayer 1 Output
1Faces in frame: 2
2Est. age: 22–30
3Primary face visible
Custom Rules
BlockSingle-person profile video
More than one face in frame
No face detected
Estimated age below 18
Face not visible throughout
Rejected
Re-record prompt sent to user
3AI moderator
Context-aware moderation across frames.
Is the face consistently visible? Are sunglasses removed partway through? Does the face match throughout the video? The AI Moderator checks what static rules cannot.
See Lasso in action
Face confidence drop
AI Moderator
Analyzing
Frame analysis
Single subject, selfie profile video, indoor
Anomaly detected
Detection confidence fluctuating frame to frame, 74% average
Pattern recognized
AR face filter — enlarged eyes and ear overlay occluding landmarks
Classification
Liveness uncertain — face partially obscured across frames
Escalated
Confidence below auto-action threshold
4Human review
Flagged videos reach moderators with full context.
The video, frame-level face data, AI Moderator assessment, and timestamps showing which frames triggered the flag. Every human decision improves the system.
See Lasso in actionQueued for review
AI Moderator reasoning
A heavy AR face filter is suppressing facial landmarks. Detection confidence stays below threshold across frames — needs a human check.
Human Review1 of 4
Detection data
Face confidence: 74% avg | Frames affected: 0:00–0:18
AI assessment
Filter obstructing detection — enlarged eyes and ear overlay suppress facial landmarks across frames
AI confidence
Face detection74%
Below auto-action threshold (90%)
This decision will train the AI for future similar content
Five capabilities included in Lasso's video face detection
Detect faces in every frame of uploaded or live video. Returns face presence, count, and position data per frame. The foundation for age estimation, counting rules, and context-aware enforcement.
Sampled frames450 frames
0:04
0:05
0:05
2 faces0:06
0:06
0:07
Per-frame faces
0:041 face97%
0:051 face96%
0:062 faces94%
0:071 face97%
0:081 face95%
Face detected
00:00:06
Frame 142 · position + count logged
Estimate apparent age range per face, per frame. Returns age ranges with confidence scores across the full video. Use for minor protection, age-gating, and compliance with DSA, OSA, and COPPA.
Gender detection comes included in video face detection results, returned alongside age range and confidence. Use it for analytics and compliance, or switch it off.
Face counting that works across an entire video. Platform rules applied to the full timeline, not individual frames. One face required throughout, or flagged automatically.
Enforce any face requirement across video frames. Full face visible throughout. No sunglasses. Liveness consistency across the recording. Describe what your platform needs and the AI Moderator checks every frame.
Sampled frames450 frames
0:04
0:05
0:05
2 faces0:06
0:06
0:07
Per-frame faces
0:041 face97%
0:051 face96%
0:062 faces94%
0:071 face97%
0:081 face95%
Face detected
00:00:06
Frame 142 · position + count logged
Lasso: next-gen AI content moderation
99.9%
On autopilot, and getting smarter every day.
Three layers of AI handle the volume, enforce your rules, and surface the grey areas. Your team sees only what needs them, and every decision they make sharpens the model.
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 video face risks. Lasso catches them.
Adult entertainment
Frame-level age verification, not just the thumbnail.
- Age verification that needs checking after the thumbnail to catch a second person.
- Platforms that verify profile photos but never scan the actual video content.
Dating
Validate that a profile video shows one real, visible person.
- Video messages where the sender's face disappears for most of the recording.
- Users speaking their contact details while showing a blank screen.
- Profile videos that cut to stock footage or scenery mid-clip to avoid face validation.
Events & live streaming
Real-time face detection across an entire live broadcast.
- Live dating video calls where one party obscures their face after the initial verification frame.
- Panelists at a virtual conference where one participant's face triggers an age flag mid-stream.
FAQs
Video face detection, answered
Every face in every frame returns a confidence score, so you see how certain each detection is rather than a single yes-or-no. Because a video gives many frames of the same face, Lasso aggregates estimates across the timeline, which makes results more reliable than a single still image. Low-confidence frames get flagged or routed to human review instead of being auto-actioned.
Image face detection analyzes one still and returns face count, age range, and gender for that single moment. Video face detection runs the same ML across every frame, so it tracks faces over time and catches mid-video changes a single image cannot, such as a second person entering frame or a mask going on partway through. A 30-second clip produces hundreds of data points that your rules and the AI Moderator evaluate as a whole.
Layer 1 extracts frames and runs ML face detection, returning count, age range, gender, and confidence per frame. Layer 2 applies your custom rules, such as face count limits and age minimums, per content type. Layer 3 is the AI Moderator, which checks context across frames, and Layer 4 sends anything uncertain to human reviewers with full context. Around 99.9% of decisions are automated, so your team only sees the edge cases.
Face detection finds faces and reads attributes like count, age range, and gender, but it does not identify who anyone is. Facial recognition matches a face against a database of known people, which is a separate capability. Video face detection stays at the detection level, so it answers how many faces and what their estimated age is, not who they belong to.
Yes. Lasso runs the same frame-by-frame detection on uploaded video and on live streams. On live content, faces are detected and classified as the stream plays, your rules are enforced immediately, and flagged content reaches moderators in seconds.
Video face detection analyzes video content frame by frame to detect and classify faces. Each frame returns face count, position, age range, gender, and confidence scores. Unlike image face detection, it works across time. A 30-second video produces hundreds of data points that your rules and AI Moderator evaluate as a whole.
Each face in each frame gets an age estimate returned as a range with a confidence score. A 30-second video might produce dozens of estimates per person. Your platform sets the rules: flag if any estimate falls below 18, require verification if confidence is low. Supports DSA, OSA, and COPPA requirements.
Lasso supports real-time face detection on live streams. Faces detected and classified as the video plays. Rules enforced immediately. Flagged content reaches moderators in seconds.
Anything you can describe. Sunglasses, masks, filters, liveness, face angle, obstruction. The AI Moderator evaluates these across the full video, not just individual frames. That means it catches mid-video changes: someone putting on a mask, another person entering frame, quality dropping. Your requirements enforced across the entire recording.
Age estimation on video content supports DSA minor protection, OSA age assurance, and COPPA compliance. Every face in every frame gets an age estimate. Below threshold: blocked automatically. Borderline cases: routed to human review. Full audit trail for compliance reporting.
See every face in your video
Lasso detects faces across every frame, estimates age, and enforces your rules automatically. It catches what a thumbnail check misses, and goes live the day you connect your first video source.
Book a demo
Lasso Moderation
© 2026. All rights reserved.
