Deepfake Detector is a web application, REST API and Chrome extension that checks whether a video, image or audio file is real or synthetic. It is run by Deepfake Detector Inc. in Montreal and was founded in 2023, after a member of the founding team received a voicemail that cloned a relative's voice.
Each check returns a verdict of Authentic, Likely Synthetic or Inconclusive, along with a confidence score and the signals behind it. The company describes itself as detection only and says it will never build synthetic-media generation tools.
How does Deepfake Detector work?
Users submit media by uploading a file, pasting a link, using the Chrome extension, or calling the API. According to the company, several forensic signal layers run in parallel against the video, image or audio, and the result is a consensus verdict rather than the output of a single model.
The result shows a verdict, a confidence score and the specific signals that drove it, such as lip-sync drift or unnatural skin texture. Low-quality input can produce an Inconclusive result, and the site advises submitting a higher-quality copy in that case.
Key features of Deepfake Detector
The service runs three detectors under one account. The video detector works frame by frame and flags failing segments, the image detector covers generators including MidJourney, DALL-E, Stable Diffusion, Flux and Firefly, and the voice detector targets cloned speech from engines such as ElevenLabs, Resemble, PlayHT and OpenAI.
Beyond detection, results can be exported as PDF, and paid plans add CSV export and detailed confidence scores. Developers get a REST API, webhooks for files over 60 seconds, and SDKs for Python, Node, Go and Ruby. Files are purged within 60 seconds of the verdict.
Key Features
AI Video Detector — Analyzes video frame by frame for lip-sync drift, generator signatures and per-frame artifacts, flagging the segments that fail. Clips can run up to 10 minutes on paid plans.
AI Image Detector — Identifies AI-generated images and faces from generators including MidJourney, DALL-E, Stable Diffusion, Flux, Firefly, Imagen, Recraft and Ideogram, with a verdict and confidence score.
AI Voice Detector — Detects synthetic speech from voice-cloning engines such as ElevenLabs, Resemble, PlayHT and OpenAI, and is built to resist added background noise.
Verdicts with confidence and signals — Each result is Authentic, Likely Synthetic or Inconclusive, with a confidence score, a TrustScore (0-100) and the specific signals behind the call. Results can be exported as PDF.
Detailed analysis and exports — Starter and above add detailed confidence scores and PDF and CSV export.
REST API — A single POST endpoint returns a JSON verdict, confidence and signals. Available on Starter and above.
Webhooks and SDKs — Webhooks return verdicts for files over 60 seconds. SDKs are available for Python, Node, Go and Ruby.
Chrome extension — Right-click any image, video or audio clip on the web to scan it. The extension is free to install.
Fast file purge — Files are processed in memory and purged within 60 seconds of the verdict. Only a hashed identifier and verdict metadata are kept for billing and audit. EU residency is available on Business plans and above.
Multilingual voice detection — The site says it can detect deepfakes in multiple languages and accents.
Pros & Cons
Pros
Covers video, image and voice detection in a single account and API
Free plan with 50 detections a month and no credit card required
Results include confidence scores and the signals behind each verdict, not just a label
Files are purged within 60 seconds of the verdict, with EU residency on Business and above
Free Chrome extension for right-click scanning, plus SDKs in four languages
Pricing is public, with 50% off Starter for verified newsrooms and non-profits
Cons
The company does not publish a specific accuracy figure, and says detection is probabilistic
The free plan limits voice and video to 2 minutes per detection
API access requires a paid plan, starting at Starter
Cancellations are not pro-rated, and paid plans queue extra detections until the next cycle
Pricing
Free $0 per month
Starter $49/mo monthly or $39/mo billed annually ($468/year) monthly or annual
Business $199/mo monthly or $159/mo billed annually ($1,908/year) monthly or annual
Enterprise $599/mo monthly or $479/mo billed annually ($5,748/year) monthly or annual
Frequently Asked Questions
What is Deepfake Detector?
Deepfake Detector is an AI tool that checks whether a video, image or audio file is real or synthetic. It returns a verdict with a confidence score and the signals behind it. It is available as a web app, REST API and Chrome extension.
How much does Deepfake Detector cost?
There is a free plan at $0. Paid plans are Starter at $49/month ($39/month billed annually), Business at $199/month ($159/month annually) and Enterprise at $599/month ($479/month annually).
Does Deepfake Detector have a free plan?
Yes. A free account includes 50 detections a month across video, image and audio, with no credit card required. Voice and video are limited to 2 minutes per detection.
What file types can Deepfake Detector check?
The site lists common video (MP4, MOV, WEBM), image (JPG, PNG, WebP, and others) and audio formats. File size and length limits depend on the plan.
Is Deepfake Detector accurate?
The company claims high accuracy across major generators but states that detection is probabilistic, not absolute. Results are confidence-scored verdicts to weigh alongside other evidence.
Does Deepfake Detector have an API?
Yes. A REST API returns JSON verdicts, with webhooks for files over 60 seconds and SDKs for Python, Node, Go and Ruby. API access starts on the Starter plan.
Are uploaded files kept private?
Files are processed in memory and purged within 60 seconds of the verdict. Only a hashed identifier and verdict metadata are retained for billing and audit. The company says it does not train on media without explicit consent.
Who is Deepfake Detector best for?
It suits newsrooms, fact-checkers, banks, insurers, fraud and KYC teams, and trust and safety teams. Individuals can also use the free plan to screen suspicious voice messages.