Free.

Free Stable Diffusion Detector. SDXL, SD3, ComfyUI, A1111.

Stable Diffusion has the strongest file-level fingerprint of any major generator. Open-source toolchains write embedded metadata by default. Our detector reads it.

1Drop imageor click to upload
2Hit checkwe analyze in seconds
3Read resultAI likelihood + source
Analysis result
0%
AI likelihood
Waiting for an image
Drop or upload an image on the left to analyze it.

The embedded metadata giveaway.

If you have an unaltered PNG straight out of Automatic1111 or ComfyUI, our detector resolves "is this Stable Diffusion?" almost instantly. The image file itself usually contains the prompt, model, seed, and sampler in an embedded text block. That block gets stripped on most platforms, after which we fall back to visual analysis.

SD3, SD3.5, and community fine-tunes.

The newer SD3 family produces subtly different outputs from SDXL. We keep coverage current as each major version lands. CivitAI-hosted community fine-tunes vary in how much they diverge from the base model, which affects accuracy on stripped images. We are honest about uncertainty when the signals are mixed.

More detection tools.

AI Image Checker AI Image Detector Midjourney DALL·E Stable Diffusion Flux Firefly

Frequently asked.

Does it work on SDXL, SD3, and SD3.5?
Yes, all major versions are covered. SDXL has the strongest detection accuracy because of its training data overlap. SD3 and SD3.5 are added as releases land.
What about ComfyUI and Automatic1111 outputs?
Both write embedded metadata by default. When the original PNG is intact, the detector reads it directly and resolves the question without running any analysis.
Will it detect CivitAI community fine-tunes?
Most are detected, with accuracy varying based on how much the fine-tune diverges from the base SDXL model. Anime, photorealism, and NSFW fine-tunes are all generally covered.