THE CASTCharacters that stay themselves
Build a character from thirteen reference fields - or drag your own art straight in - and generate a cinematic portrait via Nano Banana 2. A heraldic character sheet (front, three-quarter, back, face closeup) becomes the canonical reference that rides into every later generation, so the face on page forty is the face from page one.
Compose variants of any portrait - new pose, new view, new outfit - then compose variants of variants. Promote any of them to primary; nothing is ever lost, only archived.
THE BOOKProjects, covers, and finished exports
A project gathers cast, panels, and pages under one cover, generated with every cast portrait sent as a visual reference, primary character forward. Covers preview in memory and only commit when you say so.
Export the finished book as PDF, CBZ, or CBR - with-speech and no-speech editions, 300 DPI, auto-numbered cover-first. Whole projects round-trip losslessly between machines as a single zip.
THE BEATPanels with a memory
Generate single-beat panels at any standard aspect ratio, casting exactly the characters the moment needs - each one anchored by their sheet, so identity survives the shot. Per-panel variant overrides call on “this character, in their alt costume” without disturbing the roster.
Modify Reference mode turns any panel into the seed of the next: state what changes, what must be preserved, what gets added or erased - visual continuity inside a scene, deliberate freshness on a cut.
THE PAGEA 300 DPI page editor in the browser
An 11″×17″ canvas (spreads too) with rectangular, elliptical, and polygon regions, layered editing, alignment guides, snap, undo fifty steps deep, and a Krita-style toolbar. Drop preset grids or hand-place every region; pan and zoom each panel inside its frame.
Save Page flattens the layout server-side to a print-ready PNG. The canvas math and the renderer share one geometry spec, locked by tests; what you see is byte-for-byte what prints.
THE VOICESpeech, thought, caption, SFX
Letter bubbles onto pages or straight onto panels: draggable tails, inline text editing, per-bubble typography, bubble chaining for long lines, true cloud-silhouette thoughts, drop shadows that scale with the bubble.
Text lives in editable JSON forever; re-rendering the lettering costs nothing and calls no AI.
THE VAULTA universal asset library
Generate isolated relics, props, logos, and backdrops; extract any element out of an existing image (“the helmet”, “the carved sword”) as a new asset; remove backgrounds locally with state-of-the-art models. Everything - true assets, portraits, covers, panels - surfaces in one bridge that feeds the page editor and the panel form.
Quick Character conjures tertiary background cast in eight fields without polluting the main roster, and any of them can be promoted to a full character the day they earn it.
THE ORCHESTRATORQuick Comic: brief in, book out
A thirteen-stage pipeline - story arc, cast, style, cover, page structure, panel briefs, panels, layout, dialogue, overlays, finished PDF - with dedicated planner personas and adversarial verifiers that re-check plans before a single image is spent.
Manual, Assisted, or Auto: a server-authoritative run-ahead loop streams live progress and pauses itself the moment something needs your eyes. The demo below is this pipeline, distilled.
THE LEDGEREvery rupee accounted for
Every AI call logs its tokens and the entity it produced - permanently, even after that entity is deleted. A pricing calculator rolls the logs into per-character, per-project spend in USD and INR, deterministic and auditable against the billing console: no model ever touches the math.
The same discipline runs inside this page’s demo; the manager’s books survive any deletion.
THE VOICES BEHINDPersonas you can rewrite
Every generation is driven by a named system prompt - the Temporal Artifex renders portraits, the Schema Heraldist draws character sheets, the Sovereign Lithomancer paints covers, the Panoptic Chronomancer stages panels. All of them live as plain Markdown, editable in-app, effective on the next generation.
The demo on this page runs these exact personas, word for word.
