AI-Driven Visual Narratives
By Celso Singo Aramaki + AI (updated for 2026)
Introduction
Since 2024, generative AI has matured quickly—especially in text generation, multimodal reasoning, and production workflows. In practice, this has made AI more useful as creative infrastructure: drafting, search, revision, consistency checks, and rapid prototyping. It has also made one thing clearer: strong narrative work still depends on human direction, taste, and editing.
The Noograms Framework is an applied experiment in this space. It is not presented as a “revolution,” but as a technical, iterative attempt to use modern AI systems to support the creation of graphic novels and interactive narrative formats—while keeping authorship and creative decisions human-led.
The genesis of Noograms
Noograms started as a graphic novel project set in 1990s New York City, following an indie grunge rock band navigating early adulthood, creativity, and uncertainty. The narrative draws from graphic novel craft, long-form serialized storytelling, and literary structure—using comics as a medium for character-driven dialogue and atmosphere.
Over time, the project expanded into a framework: a set of methods and tools that help manage complexity in long-form visual storytelling (worldbuilding, continuity, pacing, character voice, and visual consistency).
What “AI-driven” means here (2026 reality)
In 2026, the most reliable use of AI in narrative work is augmentation rather than automation. The Noograms Framework uses AI where it saves time or improves clarity, for example:
- Narrative iteration: generating scene variants, testing pacing, and exploring alternative beats
- Continuity and canon support: retrieval-based systems that keep characters, locations, timeline facts, and tone consistent
- Dialogue and voice checks: stress-testing character voice across situations, then editing by hand
- Research assistance: summarizing references, comparing sources, and extracting constraints
- Production support: lightweight pipelines for prompts, style guides, and review loops—without pretending the model “understands” the story the way a reader does
Where possible, the approach favors structured inputs (scene cards, character sheets, canon notes, constraint lists) over open-ended prompting. This reduces drift and makes outputs easier to evaluate.
Technical approach (high level)
The framework treats a graphic novel as a system with maintainable components:
- Canon store: authoritative facts (timeline, locations, character rules, themes)
- Content model: chapters → sequences → scenes → panels, with metadata
- Agentic workflows (lightweight): role-based passes such as “editor,” “continuity checker,” “dialogue critic,” and “reader simulation,” each constrained by the canon store
- Evaluation loops: human review + checklists (consistency, tone, clarity, redundancy, pacing)
- Versioning: decision logs for changes, so revisions remain traceable
This is less about “AI characters” as a gimmick and more about workflow design: making long projects easier to maintain without losing artistic control.
Development process
The Noograms Framework evolves alongside the book. Tools are built, tested on real pages and scenes, and adjusted when they fail. The process is intentionally conservative:
- prefer repeatable methods over clever demos
- document decisions and constraints
- treat model outputs as drafts, not truth
- validate continuity and tone through human editorial review
Work
The Noograms Framework is an attempt to make AI useful in visual storytelling in a grounded way: clear inputs, constrained generation, traceable revisions, and human editorial control. If it succeeds, the value is practical—faster iteration, stronger continuity, and more room for artists and writers to focus on craft.