Trust & forensic methodology
How detection works, what it cannot do, how your uploads are protected, and who builds Sealed Rose.
1. How detection works
Sealed Rose uses an ensemble forensic architecture that evaluates media across multiple complementary models and spatial-frequency checks rather than relying on a single detector. For images and video frames, our pipeline combines proprietary neural artifact checks with specialized models, including detection technology from Hive, spatial frequency decomposition, and face boundary seam analysis.
For video, we sample frames across temporal segments to inspect inter-frame consistency, facial warping, and audio-video lip synchronization. For audio, acoustic formants and neural vocoder artifacts are analyzed to detect synthetic voice cloning from models like ElevenLabs and OpenAI.
2. What it cannot do
Our verdicts express mathematical probabilities, not absolute legal proof. Generative models evolve continuously, and aggressive recompression, heavy blurring, or adversarial filtering on social platforms can degrade subtle generative artifacts.
Sealed Rose provides forensic indicators and likelihood scores to help moderators, investigators, employers, and victims make informed decisions. We do not provide judicial determination, and our reports should be evaluated alongside other circumstantial evidence.
3. What happens to your file
Your privacy is fundamental to our security model. Files uploaded for verification are processed in memory and deleted after the scan completes. We do not train generative models on your uploads, and we do not sell or monetize uploaded media.
When you generate a PDF report or purchase credits, the extracted metadata and detection scores are preserved under your account so you can reference and download them. The underlying media files are discarded once frame extraction and inference finish.
4. Testing & validation
We continuously test our models against synthetic media generated by leading commercial and open-source diffusion and autoregressive architectures. This includes Sora, Kling, Hailuo, Gen-3, Veo, Seedance, Wan, Flux, Midjourney, and Stable Diffusion.
Rather than citing static marketing percentages that quickly become obsolete as new generator weights release, our evaluation focuses on sensitivity to real-world compression, social media re-encoding, and adversarial crop-and-scale attacks.
5. Who runs Sealed Rose
Sealed Rose is operated by Kinetic Helix LLC, an independent Florida company founded by Derek Gallardo. Our focus is forensic transparency, victim defense against non-consensual deepfakes and blackmail, and accessible verification tools.
If you have technical questions about our detection pipeline or security practices, reach out to our engineering team at [email protected].