Skip to main content

Introduction to RealScan

Emily Essig avatar
Written by Emily Essig
Updated over 2 weeks ago

Welcome to RealScan, Reality Defender’s web platform for detecting AI-generated or manipulated media across images, videos, audio, and text.

Whether you’re a journalist verifying footage, a trust-and-safety analyst, or a researcher testing detection models, RealScan lets you upload media and get clear, explainable results.


Create or Access Your Account

  • New to Reality Defender? Create an account at app.realitydefender.ai.
    All new users get access to our API and SDK. To import & analyze media directly in the web app, select Upgrade Plan → Analyst, or contact Sales.

  • Joining an Enterprise account? You’ll receive an email invitation to join. Follow the link to set your password and sign in.


Supported Files

RealScan formats & limits (API supports additional formats noted below)

Type

Formats

Max Size

Notes

Image

jpg, jpeg, png, gif, webp

10 MB

Use original files; avoid screenshots or heavy compression.

Video

mp4, mov

100 MB

Avoid black bars and blurry faces; one fake face can flag the whole clip.

Audio

flac, wav, mp3, m4a, aac, ogg

20 MB

Single-speaker only (≥ 9 seconds). Noisy or multi-speaker clips may return “N/A.”

Text

txt or pasted input

2,000 characters

English input recommended; see language note below.

Batch uploads: RealScan supports up to 10 files at a time.

Language support (text/audio): Production support targets English (all accents). Our methods are language-agnostic and can be extended to other languages on request.


Upload Tips for Best Results

  • Use original, uncompressed media whenever possible.

  • Keep faces clear and in focus; avoid extreme crops.

  • Don’t add black borders (pillarboxing/letterboxing).

  • For scenery or non-face content, detection is improving continuously; results may vary.


Run Your First Scan

  1. Click Upload Media or drag & drop your file.

  2. Or paste a social link from X (Twitter), YouTube, TikTok, Facebook, or Instagram.

  3. RealScan analyzes your file and displays results when processing completes.

Timing: Processing time varies by file size/quality and system load. During high traffic, scans may queue briefly.


Read Your Results

Every scan includes:

  • Risk Score (1–99%) — higher = more likely manipulated (probabilistic, not deterministic).

  • Final DeterminationAuthentic or Manipulated.

  • Per-Model Details — e.g., GAN, Diffusion, Face-Swap, Visual-Noise, or Context-Aware indicators.
    Explainability — bounding boxes and model indicators; heatmaps coming soon.

  • Download Report — export a shareable PDF or CSV.


Manage Your Workspace

  • My Scans: View your uploads.

  • Organization: For admins, manage users (change role, reset password, delete), and view company activity.
    Usage: Click your initials (top-right) to see your usage and overall company usage.


Privacy & Security

  • Files are encrypted in transit and at rest on AWS.

  • Optional delete-after-processing configuration.

  • We never train our detection models on customer data.

  • Models focus on signal/artifact analysis (not metadata content).

  • The platform maintains a range of US/UK/EU certifications and have completed SOC 2 and GDPR certification.


Need Help?


Next Steps

  • Learn to interpret scores: Understanding Results & Explainability

  • Ready to automate detection? Try the Reality Defender API

  • Need more scans or features? Upgrade to the Analyst plan for higher limits and advanced analysis.

Did this answer your question?