Verify viral political or celebrity clips
Check whether a suspicious political speech or celebrity video shows signs of face manipulation before sharing.
AI video detector
Deepfake video detector for quick review
Deepfake Video Detector focuses on human-face manipulation signals — face swaps, lip-sync edits, and identity impersonation. It analyzes facial landmarks, eye and mouth movement, lighting consistency, and skin texture to flag suspicious patterns in video clips.
Check suspicious face-swap, lip-sync, or impersonation videos with an AI likelihood score, evidence frames, and explainable signals.
Upload Your Video
Upload a clip or paste a URL to analyze the video for AI-generation signals.
MP4, MOV, WebM up to 50MB and 60 seconds on Free. Free and Starter videos are deleted within 24 hours. Results in 15–60 seconds.By scanning, you agree to the Terms and Privacy Policy.
Use cases for this check
Screen influencer and brand content
Review collaboration submissions for deepfake faces or synthetic product demos.
Check forwarded videos
Run a quick scan on videos received via DM, WhatsApp, or social media that claim to show real people.
What this check is best for
- Checking whether a person's face in a video may have been swapped or digitally altered
- Reviewing political speeches, celebrity clips, or public figures for impersonation signals
- Screening influencer or brand content before publishing or reposting
- Quick triage of forwarded videos that claim to show real people
What it can detect
- Face-swap artifacts — inconsistent skin texture, boundary blending, and facial landmark misalignment
- Lip-sync manipulation — mouth movements that do not match audio timing or speech patterns
- Lighting and shadow inconsistencies on facial features compared to the surrounding scene
- Eye and blink irregularities — unnatural gaze direction, asymmetric blinking, or fixed pupil size
- Temporal flickering — frame-to-frame jumps in face position, shape, or texture
What it cannot detect
Understanding the limits helps you set realistic expectations and combine this tool with other verification methods.
- Voice-only deepfakes or audio-only impersonation — the scanner analyzes visual signals only
- Body doubles or stand-ins where the face is not digitally altered
- Very low-resolution or heavily compressed face regions where facial detail is too degraded
- High-quality deepfakes from models that produce very few detectable artifacts
- Text overlays, subtitles, or non-face visual manipulations
Best input quality
Best results come from original video files (MP4, MOV, WebM) at 720p or higher resolution with the face clearly visible. Videos under 3 seconds, heavy platform compression (WhatsApp, Telegram), or face regions smaller than about 10% of the frame produce lower-confidence results.
Example result
This is a representative example to show what a scan report looks like. It is not a real scan result.
78
AI Likelihood Score
Verdict: Likely AI · Confidence: Medium
Signals detected:
- Lip-sync mismatch at 0:03–0:07
- Facial landmark instability at 0:12
- Lighting inconsistency on forehead vs. cheek area
Common false positives
These are situations where a real video may be incorrectly flagged as AI-generated.
- Heavily filtered videos — beauty filters, skin smoothing, and face-altering apps can mimic deepfake artifacts
- Screen recordings of video calls — low resolution and compression can create false face-swap signals
- Videos with unusual studio lighting — dramatic side-lighting or colored lighting can appear as face manipulation
- Animated or CGI characters — synthetically rendered faces may trigger face-manipulation signals
Common false negatives
These are situations where an AI-generated video may not be flagged.
- High-quality face swaps with consistent lighting and no visible boundary artifacts
- Short clips (under 3 seconds) with too few frames for reliable temporal analysis
- Videos where the face region is very small, occluded, or only partially visible
- Content that has been re-encoded multiple times, destroying subtle manipulation artifacts
Manual verification workflow
The scan is a first-pass signal. For high-stakes content, combine it with these manual steps.
- 1
Pause on evidence frames and compare the flagged facial region to earlier or later frames in the same video
- 2
Check whether the audio matches the lip movements throughout the entire clip, not just at flagged moments
- 3
Look for consistent lighting and shadows on the face versus the neck, ears, and background
- 4
Search for the original source of the video — if it comes from an unverified or anonymous account, treat with more skepticism
- 5
Cross-reference with known real appearances of the person shown in the video
How to use this detector
1
Upload the suspicious video file or paste its public URL.
2
The detector analyzes face-swap artifacts, motion, and lighting patterns.
3
Review the report with confidence score and highlighted evidence frames.
Page-specific questions
Does a high score prove it's a deepfake?
No. A high score means the scan found suspicious face-manipulation signals. It should be reviewed with source context, not treated as proof.
Can it detect all types of deepfakes?
It works best on face swaps and lip-sync edits. Voice-only fakes, body doubles, and non-face content are harder to detect. Results vary by video quality and technique used.
Does compression affect deepfake detection?
Yes. Platform compression (especially from messaging apps and social media re-uploads) can destroy the subtle artifacts detectors rely on. Upload the original file when available.
Is this the same as an AI video detector?
Deepfake detection is one type of AI video detection. This page focuses specifically on face and identity manipulation. For general AI-generated video checks, see the AI Video Detector page.
Need more scans or reports? View pricing
Ready to check a video?
Upload a clip, inspect the evidence, and decide what to do next without overstating what an automated result can prove.