# Sources And Evidence

## Web Sources

- CANVAS: Continuity-Aware Narratives via Visual Agentic Storyboarding — https://arxiv.org/html/2604.13452v1
- StoryDiffusion: Consistent Self-Attention for Long-Range Image and Video Generation — https://arxiv.org/abs/2405.01434
- Story2Board: A Training-Free Approach for Expressive Storyboard Generation — https://arxiv.org/abs/2508.09983
- AutoStudio: Crafting Consistent Subjects in Multi-turn Interactive Image Generation — https://arxiv.org/abs/2406.01388
- ViStoryBench: Comprehensive Benchmark Suite for Story Visualization — https://arxiv.org/abs/2505.24862
- StoryMem: Multi-shot Long Video Storytelling with Memory — https://arxiv.org/abs/2512.19539
- VideoMemory: Toward Consistent Video Generation via Memory Integration — https://arxiv.org/abs/2601.03655
- OneStory: Coherent Multi-Shot Video Generation with Adaptive Memory — https://arxiv.org/abs/2512.07802

## Local Code / Prompt Evidence

Current snapshot inspected: `HEAD=b9ee5f1`.

| Evidence | Path |
|---|---|
| `shot_staging` manifest entry, provider/model/schema | `backend/app/core/step_manifest.py:586-600` |
| `scene_detail` manifest entry, dependencies, schema | `backend/app/core/step_manifest.py:681-725` |
| module versions and prompt deps | `backend/app/core/version_registry.py:1-145` |
| prompt loader stem/latest semantics | `backend/app/modules/prompt_loader.py:1-140` |
| shot_staging v13 active system prompt | `prompts/_base/shot_staging/13.202605170422/system.md:1-180` |
| shot_staging v13 schema fields | `prompts/_base/shot_staging/13.202605170422/schema.json:25-69` |
| scene_detail v25 RenderPromptCard priority | `prompts/_base/scene_detail/25.202605161320/system.md:1-220` |
| RenderPromptCard field model and entry point | `backend/app/core/steps/render_prompt_card.py:1-130`, `backend/app/core/steps/render_prompt_card.py:3409-3545` |
| reference image indexing and t2i rewrite | `backend/app/services/scene_reference_service.py:34-130` |
| previous-shot background ref provenance | `backend/app/services/scene_reference_service.py:692-952` |
| state variant legacy gaze coupling | `backend/app/core/steps/image_steps.py:633-899` |
| semantic contract legacy gaze coupling | `backend/app/modules/semantic_contract_router.py:1-130` |
| shot_visibility gaze_target legacy path | `backend/app/modules/pipeline/shot_visibility.py:310-420` |

## Notes

- Scope is whole-project architecture direction, not a narrow wave review. The code anchors above are representative production paths for the pipeline, prompt, contract, reference, memory, and validation layers.
- This is not a line-by-line audit of every file in the repository. It is a research-to-architecture analysis grounded in current code and active prompt packs.
- The active repository also contains in-progress Area #2 state/gaze separation work. Recommendations assume that cleanup should finish before introducing the story memory layer.