# WhatMadeIt

> WhatMadeIt is a privacy-first AI content authenticity and digital forensics platform. It detects AI generation signals across images, video frames, audio clips, and documents, while inspecting camera EXIF metadata and C2PA Content Credentials with transparent, verifiable evidence.

## Core Principles & Capabilities
- [In-Browser Privacy](https://whatmadeit.com/): Original video, audio, PDF, and Office files are processed locally inside the browser. Sensitive user files are never uploaded to remote servers.
- [Multi-Signal Forensics](https://whatmadeit.com/methodology): Instead of opaque confidence scores, WhatMadeIt separates visual diffusion texture anomalies, camera EXIF hardware parameters, and cryptographic C2PA provenance into independent inspectable signals.
- [Evidence Limitations](https://whatmadeit.com/limitations): Openly explains the boundaries of AI detection, including false positives, social media metadata stripping, and why absence of credentials proves nothing.

## Free Browser Tools
- [Image Metadata & EXIF Checker](https://whatmadeit.com/tools/image-metadata-checker): Inspect camera model, lens settings, ISO, shutter speed, creation dates, and editing software tags locally without uploading files.
- [C2PA Content Credentials Viewer](https://whatmadeit.com/tools/c2pa-viewer): Scan JPEG, PNG, and WebP files locally for C2PA container markers and JUMBF boxes.
- [Content Credentials Marker Checker](https://whatmadeit.com/tools/content-credentials-checker): Quick browser-based check for embedded C2PA authenticity manifests.
- [PDF Authenticity Checker](https://whatmadeit.com/tools/pdf-authenticity-checker): Inspect PDF structure, creator software names, revision history, and embedded markers locally.
- [Video Frame Checker](https://whatmadeit.com/tools/ai-video-detector): Extract and evaluate video frames locally in the browser.

## Authoritative Forensics & Detection Guides
- [How to Read C2PA Content Credentials](https://whatmadeit.com/guides/how-to-read-c2pa-content-credentials): In-depth engineering breakdown of C2PA manifest stores, cryptographic assertions, and signing implementations across OpenAI, Adobe, and Leica hardware, plus the "analog hole" re-photography limitation.
- [How to Tell If an Image Is AI-Generated](https://whatmadeit.com/guides/how-to-tell-if-an-image-is-ai-generated): Comprehensive 5-step framework comparing visual cues, file metadata, and provenance signals.
- [How to Detect Midjourney Images](https://whatmadeit.com/guides/how-to-detect-midjourney-images): Guide to identifying Midjourney v6/v5 signatures: waxy skin sheen, volumetric lighting, coherence anomalies, and prompt drift.
- [How to Detect FLUX.1 Images](https://whatmadeit.com/guides/how-to-detect-flux-ai-images): Guide to photoreal FLUX.1 outputs: depth-of-field edge blending, micro-skin pores, typography limits, and ComfyUI metadata.
- [How to Detect DALL-E 3 Images](https://whatmadeit.com/guides/how-to-detect-dall-e-3-images): Identifying ChatGPT/DALL-E 3 creations: hyper-saturated color palettes, illustrative gloss, and OpenAI-signed C2PA credentials.
- [How to Interpret AI Detector Scores](https://whatmadeit.com/guides/how-to-interpret-ai-detector-scores): Explains why AI detector scores are not probability calculations and when uncertainty is the correct verdict.
- [Content Credentials Explained](https://whatmadeit.com/guides/content-credentials-explained): What signed records show, how certificates work, and why stripped metadata does not imply artificial generation.

## Pricing & Access
- [Free Tier](https://whatmadeit.com/pricing): 3 free daily checks with no sign-up or credit card required.
- [Plus Plan](https://whatmadeit.com/pricing): 100 checks/month, priority processing, consumer summary exports.
- [Pro Plan](https://whatmadeit.com/pricing): Unlimited checks, deep professional forensic reports, print/PDF export, full signal breakdowns.

## Important Interpretation Rule
An assessment describes technical signals found in a file. It does not prove authorship, intent, copyright ownership or legal admissibility. A missing metadata field or missing Content Credentials manifest is not, by itself, evidence of AI generation.