Current development stage
What is live now, what we are refining, and what is next.
Adaptive media tuning is live
Image and video watermark strength is tuned per region to stay invisible and still recover reliably.
Audio tuning by segment
Audio watermarking is tuned by frequency band and time window, not one fixed pattern.
Dual-path verification
If C2PA metadata exists, we use it. If metadata is stripped, UID recovery and decode logs still support verification.
Deep Scan fallback is improving
When UID recovery is weak, Deep Scan uses logs and signal traits to return an attribution confidence score.
Text/docs/code preview started early
Text/document/code watermark preview sections are now active in the product showcase ahead of the original plan.
No-watermark AI-origin is in development
DeadDetect Deep Scan now includes active model development for AI-origin estimation with fingerprint-style evidence output.
Current build focus
- Calibrate watermark strength per media region for invisibility + recoverability
- Expand transform testing (re-encode, resize, crop, compression, clipping, noise)
- Improve Deep Scan scoring when direct UID recovery is degraded
- Standardize operator evidence output for image points, video events, and audio traces
- Harden text/document/code verification for messaging and light reformat scenarios
- Build AI-origin fingerprint confidence output for no-watermark Deep Scan
Stage map
Core media baseline
Watermark embed/detect is active for image, video, and audio with measurable verification outputs.
SDK/API hardening
Improve decode reliability, logging, survivability reporting, and verification flows for pilots.
Format + model expansion
Text/document/code support is active and AI-origin estimation for non-watermarked content is in development.
Near-term note
Site visuals are demo overlays for operator review. Core delivery remains SDK/API workflows for backend verification, moderation, and audit pipelines.