Stability AI
Stable Diffusion open models
What Stability AI says Stable Diffusion open models puts into the files it makes, by file type, quoted from Stability AI's own pages. Covers Stability AI's generative-models reference code (SDXL, SDXL-Turbo and Stable Video models, including SV3D and SV4D) that people run themselves, not its hosted services.
At a glance
| File type | Content Credentials | Invisible watermark | AI label in metadata | Visible mark | Tool named in metadata |
|---|---|---|---|---|---|
| Images | Not stated | Yes (reference code) | Not stated | Not stated | Not stated |
Images
- Content CredentialsA signed record (C2PA) of how the file was made, sealed so any change shows.
Not stated. None of Stability AI's pages listed under Sources read says so.
Stability AI does not say that its open models add Content Credentials.
- Invisible watermarkA signal woven into the pixels, frames, audio or wording, meant to survive copying.
Stability AI's generative-models repository says images generated with its code embed an invisible watermark using the invisible-watermark library, and that this watermark differs from the one in Stable Diffusion 1.x and 2.x.
This describes the reference code, which users run and can edit themselves. An undated Stability AI page that appears to date from 2024 says watermarking is included by default in its open image models.
generative-models README (Stability-AI GitHub)
Quoted passage Invisible Watermark Detection
Images generated with our code use the [invisible-watermark](https://github.com/ShieldMnt/invisible-watermark/) library to embed an invisible watermark into the model output. We also provide a script to easily detect that watermark. Please note that this watermark is not the same as in previous Stable Diffusion 1.x/2.x versions.
Artificial Intelligence and the Content Ecosystem
Quoted passage Paragraph on safeguards ('We apply an imperceptible watermark …')
We apply an imperceptible watermark to images generated on our platform, and include watermarking by default in our open image models." "With significant elections in over 60 countries impacting 4 billion people this year, we believe safeguarding democracy is of paramount importance.
- AI label in metadataA plain metadata field (IPTC Digital Source Type) saying AI made the content. Anyone can add or remove it.
Not stated. None of Stability AI's pages listed under Sources read says so.
- Visible markA mark you can see, such as a logo or a sparkle in a corner.
Not stated. None of Stability AI's pages listed under Sources read says so.
- Tool named in metadataThe tool or model written into ordinary file fields, such as Software or Creator.
Not stated. None of Stability AI's pages listed under Sources read says so.
Limits and detection
- Checking tool from the makerA way the maker offers to check a file for its traces.
Stability AI's generative-models repository provides a script to detect the invisible watermark in images generated with its code.
generative-models README (Stability-AI GitHub)
Quoted passage Invisible Watermark Detection
Images generated with our code use the [invisible-watermark](https://github.com/ShieldMnt/invisible-watermark/) library to embed an invisible watermark into the model output. We also provide a script to easily detect that watermark. Please note that this watermark is not the same as in previous Stable Diffusion 1.x/2.x versions.
Sources read
Each page below was read and the quoted passage found on it. Last verified October 19, 2026.
- Stable SafetyStability AI
- Artificial Intelligence and the Content EcosystemStability AI
- Release Notes (Stability AI API Platform)Stability AI
- generative-models README (Stability-AI GitHub)Stability AI
- Terms of Service (Stability AI)Stability AI
