Dual-Model Vision Ensemble · Heatmap Analysis

AI Image Detector

Upload any photo or illustration to verify authenticity. Combines high-resolution Vision Transformers with embedding towers and localized AI heatmaps.

Upload Image JPG, PNG, WEBP · up to 10 MB

Drag & drop an image here, or click to browse

Supports ChatGPT, Midjourney, Flux, Stable Diffusion, and real photography.

No image selected

AI / Fake --

Real / Camera --

AI Confidence: 0% Real Confidence: 0%

Visual Inspection & Detection Heatmap

Toggle between original image and the neural heatmap overlay.

Heatmap Scale:

Real (0%) → Synthetic AI (100%)

Warmer red & orange regions pinpoint artificial generator tokens and diffusion textures.

Multi-Model Vision Ensemble Breakdown

Our detector executes multiple specialized computer vision architectures in parallel to eliminate single-model blind spots.

Vision Transformer (ViT-Deploy)

--%

Trained on generator tokens, patch frequency artifacts, and latent diffusion textures (DALL-E 3, Midjourney, Flux, SDXL).

Smaug AI Image Detector

--%

Gemma-based vision embedding classifier verifying global perceptual semantics and latent coherence.

Verification Notice: AI detection provides probabilistic evidence. New generator releases, extreme compression, or heavy manual filters can alter sensor signatures. Always consider visual context when making critical determinations.

State-of-the-Art Ensemble

Combines ViT token classification with Smaug vision towers to detect DALL-E, Midjourney, Flux, and SDXL.

Private by Design

Images are analyzed directly in server memory and never stored or shared.

Developer API

Integrate automated image scanning via POST /api/v1/detect-image with your API key.