UnmarkPro
Free · Finds which AI made it

What Made This? AI Image & Video Checker

Drop any image or video to discover which model generated it and what marks it carries. Inspects Content Credentials (C2PA), camera EXIF, generator prompts, and encoder tags locally with zero uploads.

Choose an image or video to check

Supports JPEG, PNG, WebP, MP4, and WebM • 100% private, stays on your device

How In-Browser AI Provenance Detection Works

Understanding what signals AI models leave inside files, and what can and cannot be verified honestly.

Content Credentials (C2PA)

Leading AI generators cryptographically sign output files with C2PA manifests. Reading these containers gives verified proof of the generator and signing authority.

Hardware Camera Profiles

Authentic photographs contain physical lens, sensor, and shutter telemetry in EXIF headers. Genuine camera captures are verified by matching physical hardware profiles.

Embedded Prompt Chunks

Tools like Stable Diffusion and ComfyUI embed generation parameters and node graphs in image headers. We extract these parameters and let you sanitize them.

Honest Detection Limits

If a file has been re-saved through a screenshot, edited in software that strips metadata, or shared through social platforms that discard headers, provenance metadata is lost. When evidence is not present, we report that the model cannot be determined. We never invent unmeasured certainty scores.

Frequently Asked Questions

How do I check which AI model made an image or video?

Open this page, drop your image or video into the inspection area, and read the plain-language report. The tool analyzes Content Credentials (C2PA) manifests, camera EXIF tags, generator prompts, and encoder signatures locally in your browser. Note: if a file has had all metadata stripped by a screenshot or social platform, we cannot identify the generator.

What is a C2PA Content Credentials manifest?

Content Credentials (C2PA) is an open cryptographic standard used by major AI platforms — including OpenAI (Sora, DALL·E), Google (Veo, Imagen), Adobe (Firefly), and Microsoft — to sign media with details about how and when it was created. Social networks read these manifests to attach automated AI notices.

Why does the detector report 'Could not determine source model' for some files?

We only report a source model when cryptographic C2PA credentials, generator prompts, or verified metadata signatures are genuinely present. We do not show unmeasured guesses or invent confidence percentages. When a file has been stripped of metadata or re-saved without provenance tags, the honest verdict is that the model cannot be determined.

Does this detector upload my private files to any server?

No. All inspection runs entirely in your browser using local WebAssembly and JavaScript. Your image or video never leaves your device, and closing the browser tab purges temporary session data immediately.

Can I remove watermarks or C2PA tags after detection?

Yes. When the detector identifies a visible watermark or Content Credentials manifest, it provides a direct step to clean the watermark or download a sanitized copy with all provenance metadata stripped.