DaVinci Text-Stability Video Pipeline

Protocolverbatimlex:davinci-text-stability-video-pipeline

Three-phase pipeline solving stochastic glyphic degradation in generative video (text melting in AI video). Phase 1 (Token Lock): Generate perfect static frame via Ideogram 2.0 (T5 text encoder treats text as geometric shapes). Phase 2 (Temporal Injection): Feed anchor frame into Image-to-Video model (Runway Gen-3 Alpha or Luma Dream Machine) as first-frame conditioning — model moves existing pixels instead of hallucinating new ones. Phase 3 (Sub-Pixel Reconstruction): Upscale via Topaz Video AI Proteus model to 4K/8K at 60fps. Core principle: Never ask a video model to WRITE. Ask it to MOVE what is already written.

Quoted verbatim from Notion, captured 2026-07-25. First seen 2026-03-29.

Provenance

Every source that names this term. Links into the Notion workspace require workspace access.

Notionapp.notion.com/p/332a981d75b681eebc6fe5480fdbb635 (private workspace)2026-07-25

Cite

DaVinci Text-Stability Video Pipeline — lex:davinci-text-stability-video-pipeline — Neuruh Lexicon v1.0.0 (0a2bb72426e9) — https://lexicon.neuruh.com/t/davinci-text-stability-video-pipeline

Agents: GET /api/term/lex:davinci-text-stability-video-pipeline or the lexicon_resolve MCP tool.