Voice consistency
81%
- Prosody smoothingHigh
- Noise continuityMedium
- Speaker driftFlagged
Working browser prototype
Analyze suspicious media in real time.
This prototype accepts uploaded images, audio, and video, generates an explainable authenticity score, and includes a live camera simulation for scam-call style checks. It is built as a static site, so you can publish it quickly on Netlify, Vercel, or GitHub Pages.
Upload media
Preview and inspect image, video, and audio directly in the browser.
Explain verdicts
Returns signal bars, reasons, next-step guidance, and an exportable report.
Deploy fast
Single static site with no backend requirement for the prototype version.
Image artifacts
89%
- Boundary blendFlagged
- Light mismatchFlagged
- Texture spikeMedium
Video timeline
76%
- Lip-sync gap220 ms
- Frame burst11 frames
- Identity stabilityDrop
Prototype output
Explainable
Upload a real file, inspect the preview, run analysis, and export a browser-generated summary that can be shown in a demo or pitch.
Prototype analyzer
![Media preview]()
Upload a file and get a verdict.
This prototype runs fully in the browser. It extracts visible metadata, simulates detector signals by modality, and produces a realistic trust-and-safety review flow with confidence, evidence, and export.
1. Choose how to test
Use your own file or load one of the built-in demo samples to show the product flow quickly.
Demo tips
For pitch demos, use one real-looking sample and one suspicious sample. The exported report helps you show an operational moderation flow even without a backend.
2. Preview and scan overlay
Preview updates automatically when a file or sample is loaded.
Load a file or a sample to start the authenticity review.
Authenticity risk
--
Confidence will appear here
Waiting for analysis
Explainability output
Run an analysis to populate reasons, suspicious signals, and review guidance.
Signal summary
- Detector stateIdle
- Last exportNot generated
Run analysis first
What this prototype does
Shows the UX and decision flow of a deepfake detector: ingest, preview, score, explain, and export.
What it does not do yet
It does not call production ML models. The scores are generated from deterministic browser-side heuristics for demo use.
Why this is still useful
You can publish it immediately, demo the product clearly, and later replace the scoring logic with real APIs or on-device models.
Live camera mode
Simulate a real-time call check.
Start the camera feed and the prototype will generate rolling risk updates like a live call authenticity assistant.
Start live mode to simulate a scam-call monitoring flow. If camera permission is denied, the prototype will still run a simulated timeline.
--
Current live risk
0
Rolling windows scanned
0
Suspicious spikes
Live review timeline
Each rolling window produces a short explanation that a moderator or end user can read.
Live mode idle
Start the camera feed to generate rolling analysis updates.
Recent windows
- StatusNot started
Ready to publish
Put it online as a static site.
Because this prototype is a single static website, publishing is simple. Upload the files to a static host and it will work immediately.
Netlify
Drag the project folder into Netlify Drop, or connect a Git repository. The `index.html` file becomes your home page automatically.
Vercel
Import the folder or repository as a static project. No framework settings are required for this version.
GitHub Pages
Push the files to a repository and enable Pages on the root branch. This works well if you want a free public demo link.
Current deliverables
`index.html` contains the full working prototype, and `README.md` explains fast publishing steps.
Next upgrade path
Replace the browser-generated scores with a real backend API for voice, image, and video inference.
Best demo flow
Show one uploaded suspicious file, one live-mode run, then export the JSON report to prove an operational workflow.