The Auto Shot Creator: Narrative Orchestration & Context-Aware Sequencing
In high-end pre-visualization, the transition from a single concept to a timed, 30-shot sequence represents a significant production bottleneck. The Auto Shot Creator is a dual-page workspace within Ploty designed to automate the generation of structured, timed, and contextually aware storyboards.
The module architecture manages complex multi-stage state machines, implements recursive LLM prompt optimization, and maintains non-destructive data synchronization between local SQLite databases and generative AI pipelines.
Technical Architecture: The Two-Stage Generation Pipeline
The Auto Shot Creator operates as a stateful workflow that decouples narrative ideation from final asset execution. This ensures surgical control over sequence parameters before data is committed to the project.
- Stage 1: The Narrative Blueprint: Managed by a dedicated Zustand slice, this stage handles high-level parameters including duration, target positioning, and prompt optimization. An LLM-Tuning Layer intercepts user input and expands it into structurally sound cinematic instructions, ensuring appropriate pacing and scene composition.
- Stage 2: The Review Draft Workspace: This stage instantiates a dynamic "Draft Mode," allowing for the manipulation of LLM-generated shot cards. Users can reorder, expand, or edit voiceovers and timing before committing to the final database write. This "Draft-before-Commit" pattern prevents database pollution and facilitates rapid scenario testing.
- Context-Aware Story Logic: A "Temporal Look-Ahead/Look-Back" algorithm analyzes the five shots preceding and following the target insertion point. This context is injected into the LLM prompt, ensuring the new sequence functions as a logically and visually consistent bridge within the existing narrative.
Key Feature Breakdown
The Shot Creator provides tools for both initial ideation and surgical project updates:
| Tool | Technical Implementation | Creative Purpose |
|---|---|---|
| Surgical Overwrite Engine | Granular boolean toggles (Image, Desc, VO) for existing records. | Updates specific storyboard data without losing historical work or deleting shots. |
| Asynchronous Execution | A Rust-based worker pool for parallel image generation. | Builds a 30-shot sequence in seconds, automating the path from text to a fully-formed board. |
| Actor Token Preservation | Programmatic injection of actor tokens into the optimization chain. | Ensures custom-defined characters remain visually and narratively consistent across drafts. |
| Non-Destructive Layering | Logic that archives old images to a background layer instead of deleting. | Preserves visual history, allowing for the retrieval of previous iterations. |
| Temporal Timing Calc | Mathematical distribution of project duration across shot lengths. | Automatically manages pacing to ensure perfectly timed boards for a specified duration. |
Performance and Optimization
High-efficiency data handling ensures that generating large sequences remains stable within the UI and database environments:
- Snapshot-Driven History: Before executing an "Accept and Execute" command, the system pushes a global snapshot to the Undo/Redo stack. This facilitates a total revert of an entire 20-shot generation in a single action, ensuring project safety.
- Optimized Draft Reloading: A
Load Draftfeature serializes the JSON generation history to the local SQLite database. This allows previous narrative concepts to be revisited and rerun with different style settings without re-entering prompts. - Database Cycle-Through: Using a custom keyboard hook, the system facilitates rapid traversal of the
description_historyandprompt_history. This feature pulls directly from the database to allow for the real-time comparison of iterations.
Core Architectural Benefits
The Auto Shot Creator functions as a Narrative IDE, integrating cinematic logic with automated workflows.
- Continuity-First Engineering: By incorporating context-awareness, the system addresses the common issue of disjointed scenes in generative storytelling. The architecture evaluates preceding and succeeding content to create a seamless narrative flow.
- Surgical Precision: The ability to overwrite specific data types across a sequence provides the control necessary for iterative professional feedback loops.
- Automated Production Pacing: Linking total duration to individual shot timers ensures that the resulting storyboard is immediately compatible with the Animation page and export requirements.
- Resilient Project Persistence: The Tauri and SQLite stack ensures that complex multi-shot generations are handled with transactional integrity, maintaining lean and portable local project files.