How OiiOii Works

1. Art Director Agent

The Art Director sets the overall visual language: target style (anime, cinematic, 3D, watercolor, etc.), color palette, lighting mood, and aspect ratio. Every downstream agent inherits these constraints, which is what guarantees stylistic consistency across the entire short — something single-model AI video tools cannot do because they regenerate style decisions per clip.

The Art Director also picks the rendering backend per shot. OiiOii has access to 25+ underlying models — Sora 2, Veo 3.1, Seedance 2.0 / 2.6, DreamActor-M1, Kling, Kling o1, Hailuo 02, Vidu, Midjourney, Seedream 4.0 / 4.5 / 5.0, Nano Banana, NovelAI — and the Art Director selects the best one for each shot based on style and motion needs.

2. Scriptwriter Agent

The Scriptwriter expands a one-line idea into a structured screenplay: scene list, character beats, dialogue, voice-over, and on-screen action. It understands narrative arcs (setup, conflict, payoff) and can produce shorts in formats ranging from 15-second hooks to 60+ second narrative pieces.

If you provide your own script, the Scriptwriter parses it into the same internal scene graph so all downstream agents can work from the same source of truth.

3. Character Designer Agent

This is the agent that solves the face-drift problem. Once the Character Designer locks in a character — face structure, outfit, proportions, hair, signature accessories — those identifiers are persisted and passed to every shot in which the character appears.

Characters can be created from a text description or from a reference image. The Character Designer produces a multi-pose character sheet (front / 3-4 / side / back, multiple expressions) that downstream agents reference to keep the character recognizable across hundreds of shots.

4. Scene Creator Agent

The Scene Creator builds reusable scene assets: a forest, a classroom, a spaceship interior — each captured as a consistent environment that can be re-shot from different angles without lighting or prop drift.

Scenes carry metadata for time of day, weather, mood, and key props. When the Animator agent generates a new shot inside that scene, the Scene Creator passes those metadata so the new shot fits visually.

5. Animator Agent

The Animator turns the storyboard into actual moving video. It orchestrates the chosen rendering backend (Sora 2 / Veo 3.1 / Kling, etc.), passing the character identity from the Character Designer, the scene environment from the Scene Creator, and the action description from the Scriptwriter.

The Animator handles motion continuity across shots — a character walking left in shot 3 will keep walking left in shot 4, and the camera language stays consistent.

6. Editor Agent

The Editor sequences the shots into a coherent timeline, handles transitions, paces beats according to the script, and trims clips for narrative impact. It can apply standard editing patterns (match cut, J/L cut, montage) and reorders shots if pacing benefits.

All edits are non-destructive. You can regenerate a single shot via the canvas and the Editor reflows the timeline automatically.

7. Sound Engineer Agent

The Sound Engineer generates background music tuned to the story's emotional arc, plus diegetic sound effects (footsteps, ambience, impact) matched to on-screen action.

You can bring your own music. The Sound Engineer handles mixing levels, ducking under dialogue, and matching cuts to musical beats.

FAQ