Media authenticity standard
Verifying authentic human media in an age of generative AI.
In short
Sealed Rose establishes media authenticity by combining detection partner foundation models with our own multi-frame temporal and optical flow checks. Synthetic AI clones generated by Wan 2.1, Kling, Sora, Runway, and ElevenLabs are measured against the benchmarks on our methodology page.
Frequently asked questions
How does Sealed Rose distinguish between real video and an AI deepfake?
We combine detection partner foundation models with our own multi-frame temporal and optical flow checks across 10–45 sampled frames. Real video exhibits natural optical sensor noise and consistent physiological motion, while generative AI video models (Wan 2.1, Kling, Sora) leave latent diffusion grid artifacts, temporal flickering across frame boundaries, and phase inconsistencies.
Can deepfake detection identify cloned voices in audio?
Our acoustic engine performs Short-Time Fourier Transform (STFT) spectral analysis, detecting high-frequency phase discontinuities and robotic vocoder artifacts characteristic of neural voice synthesis (ElevenLabs, Cartesia).
What is container and EXIF metadata verification?
Digital cameras and recording software write standardized container tags and EXIF data upon capture. Sealed Rose inspects media files for consistent metadata and flags stripped or synthetic container profiles.
Are uploaded videos and photos stored or trained on?
Sealed Rose utilizes an ephemeral RAM-only architecture. Media is buffered exclusively in volatile memory during analysis and deleted right after your scan once the forensic report is generated. No user media is ever retained or used for model training, except images you send to search engines, kept up to 1 hour.
Can I verify images and videos for free?
Sealed Rose offers free online verification tools at /verify-video, /verify-image, /verify-audio, and /verify-text with 3 complimentary checks per day and no account required.