Choose an image to see saved software names, camera information and AI prompts or settings. What you can read depends on the information still in the file.
Drop an image here, or choose a file above
PNG, JPEG or WebP / one file, up to 32 MiB
Your image is read in your browser. This tool does not send your image, filename or results to a server or save them automatically. If you use an external verification service, check how it handles your data.
Choose an image to see the information we can read here.
Metadata can be copied or edited. A mention is not proof of AI generation or of the named provider. Missing information does not rule out AI generation.
Settings saved in the image are shown up to 4,000 characters per value. These records can be edited later. For ComfyUI, we read named values from supported nodes and show each node separately. We do not guess settings from unsupported nodes, connections or values without field names.
Prompts, creator names and location data may appear as saved in the file. Personal information is not hidden automatically. Each value is shown up to 4,000 characters. We only look for sections that may hold C2PA records; we do not read their contents or editing history.
The saved file includes the image filename and readable information without hiding personal details. Check for creator names, location data and other private information before sharing it.
Use the original image file when possible, as it is more likely to retain its records. Copies saved from social media and screenshots may have lost this information. Results show the records and where they were found. A service that verifies C2PA signatures can help you check creation and editing history.
Records can describe providers, editing actions and AI use. C2PA also supports camera-captured photos. A container alone does not establish AI generation; signatures, asset integrity and generation actions need validation.
Product names, Software fields, generation parameters and ComfyUI workflows can be clues. Their presence depends on settings and export paths. A product name in a prompt can also match, so these results are advisory.
Technologies such as SynthID embed information in the image itself. They need a compatible detector and are not checked here. They may exist even when file metadata is absent.
Detection models estimate origin from image features, but accuracy changes with unfamiliar generators and editing. Visible logos and publisher disclosures are other clues. This page does not automatically check these signals.
Reads PNG tEXt, zTXt, iTXt, eXIf and caBX; selected JPEG EXIF fields, standard XMP, comments and some APP11; and WebP EXIF, XMP and C2PA. Extended JPEG XMP, full fragmented JUMBF reconstruction, external provenance recovery and signature verification are not supported. Limits: 2 MiB per metadata region, 8 MiB total extracted metadata, 128 fields. HEIC, AVIF and GIF are not supported. Full pixel-stream integrity is not checked.
Searches readable metadata for names including OpenAI, Google / Gemini / Imagen, Adobe Firefly, Midjourney, Stable Diffusion, ComfyUI and FLUX. This does not mean every image from every provider can be detected.
To calculate a text hash, use the hash generator