DeadMark
Invisible watermarking for image, video, and audio - C2PA-aware, log-backed, and built to survive real-world transforms.
The goal is simple: keep it invisible, keep it verifiable, and keep it alive after edits.
For image and video we tune changes as close to pixel level as possible while balancing invisibility, robustness, and survivability. For audio (including video audio tracks) we apply subtle edits at selected time-frequency instances to keep the watermark inaudible but recoverable.
Every encode produces structured logs (CSV/JSON) to support audits and later model training. It can also work alongside C2PA manifests when they are present in your distribution workflow.
DeadMark uses an AI-driven, ML-adaptive multi-carrier stack across image, video, and audio. The system automatically decides the best technique mix and strength for each asset so watermark signals stay as invisible as possible while maintaining robust verification after common edits.
C2PA + DeadMark workflow
We use C2PA manifests when present, but DeadMark does not depend on C2PA alone. Our primary verification path is the invisible UID watermark + internal decode logs, so source tracing still works even if a platform strips manifest metadata.
In practice, our UID recovery is designed to remain useful even when manifest recovery fails. C2PA strengthens provenance. DeadMark keeps verification continuity.
What ships together
- Invisible media watermark (image / video / audio)
- UID-linked encode + decode logs
- Detection overlays for operator review (product UI)
- C2PA manifest support in provenance workflows
- Fallback verification when C2PA is stripped
C2PA when available. DeadMark verification even when it is not.
We support standards-based provenance (including C2PA), but verification does not depend on a manifest surviving the full content journey. DeadMark uses an embedded UID signal plus timestamped internal logs to preserve traceability.
In real distribution paths, metadata is often lost before the embedded signal. That is why we treat C2PA as a valuable layer, not the only layer.
Generate / ingest media
Media enters the DeadMark workflow through our operator pipeline. C2PA can be attached when present, but generation and verification do not depend on it.
Embed UID + internal logs
DeadMark embeds a source-linked UID signal and records decode/verification context in internal logs. This keeps continuity even when external metadata is stripped.
C2PA provenance layer
C2PA manifests are additive and useful for standards-based provenance. We surface them in review output when available, but they are not the only proof path.
Recover, verify, and audit
Verification uses embedded UID recovery, decode overlays, and timestamped internal logs. If C2PA is missing, the embedded signal and logs still support traceability.
Why we show both C2PA and internal evidence
Standards help interoperability. Embedded signals and internal logs add resilience. Showing both gives a complete verification path without exposing proprietary internals.
Roadmap note
Text/document/code support was started early and is now visible in public preview UI. In parallel, DeadDetect no-watermark AI-origin estimation is in active development.
Text, docs, and code watermarking
DeadMark is expanding beyond image/video/audio with one API and format-aware profiles for text workflows. The target is practical survivability through copy/paste, spacing changes, and light reformatting.
export function scoreRisk(input: string): number {
const normalized = normalizeText(input);
const entropy = estimateEntropy(normalized);
return Math.min(100, Math.round(entropy * 37));
}