Why it exists
Visible AI badges, provenance metadata, and embedded watermark signals are different technical layers. Delete SynthID provides one workflow for supported images while explaining those differences and avoiding a universal success claim.
How the service is built
After Google sign-in, an upload is stored in private Cloudflare R2 and queued in PostgreSQL. A remote AWS Spot CPU worker claims the job. The actual image processor runs in a read-only Docker container with networking disabled, using preloaded software and models. The result returns to private R2 and is available through an authenticated account download.
This is server-side processing, not on-device processing. The processing container is offline while it runs, but the surrounding host communicates with the queue and private storage.
Product principles
- Describe the architecture and present limitations plainly.
- Require users to own or have permission to modify their images.
- Use one-time credits instead of a recurring subscription.
- Keep image access tied to the authenticated account workflow.
Independent service
Delete SynthID is an independent service. It is not affiliated with, endorsed by, sponsored by, or operated by Google or Google DeepMind. “Google,” “Google DeepMind,” “Gemini,” and “SynthID” are names or marks of their respective owners and are used only to identify relevant technologies.
What we do not promise
We do not promise perfect reconstruction, complete signal removal, a negative result from a proprietary detector, uninterrupted availability, or suitability for every image. Users should inspect their output and preserve their originals.