Video.
Sovereign text-to-video, image-to-video, and editing at frontier quality on operator hardware, with per-frame attestation and liveness nonces.
Video is the kernel video brain. It runs Aurora, Mickai's sovereign text-to-video and image-to-video generation engine, at quality matching the closed frontier models (14B+ parameter video-diffusion and DiT architectures). Aurora handles long-form generation, motion control, camera trajectory, multi-shot consistency, and avatar animation. Lantern handles editing: cuts, transitions, colour grading, frame-level inpainting, and temporal compositing. Every frame is signed in-band with a hardware-attested key and a liveness nonce derived from the current session (patent 09). A receiver can verify in real time that an avatar is genuine, owned by the claimed person, and not a replay. This is the architectural answer to the deepfake problem. Combined with Wax-Seal per-frame attestation and the Mickai audit ledger, every Mickai video is regulator-verifiable down to the frame. Generation runs entirely on operator-controlled silicon.
- 01Sovereign text-to-video and image-to-video generation at frontier quality
- 02Long-form generation with motion control, camera trajectory, multi-shot consistency
- 03Frame-accurate editing: cuts, transitions, colour grading, inpainting
- 04Per-frame attested avatar rendering with liveness-nonce embedding
- 05Real-time receiver verification interfaces (anti-deepfake architecture)
- 06Signed per-frame provenance into the audit ledger
Authoritative external corpora and standards this brain treats as canonical. Every retrieval against these sources is signed into the audit ledger so a regulator can prove which evidence drove which output.
- 01Mickai Patents 07, 08, 09, 11, 16
- 02Frontier text-to-video and image-to-video architecture research
- 03Long-form video-diffusion-transformer (video DiT) research
- 04Open-weights video-diffusion research corpus
- 05Efficient video-generation architecture research
- 06Camera-trajectory and motion-control conditioning research
- 07C2PA video provenance specification
- 08H.264, H.265 (HEVC), H.266 (VVC), AV1 codec specifications
- 09MP4 and Matroska container formats
- 10BBC R&D video research archive
- 11ITU-R BT.709 and BT.2020 colour spaces
- 12Liveness-detection biometrics literature
Mickai-native tooling primitives this brain implements internally. Codex for sovereign plain-text graph PKM, Lectern for spaced-repetition memory, Stele for citation-provenance, and domain-native primitives layered on top. No external services in the trust path; data stays on operator-personalised hardware.
- 01Aurora (sovereign video generation engine)
- 02Lantern (frame-accurate video editing surface)
- 03Marble (storyboard canvas)
- 04Cataloguer (clip and frame management)
- 05Wax-Seal (per-frame attestation signing)