Aurora — AI Creative Production System
A JSON-based creative production system connecting strategy, structured direction and human-reviewed generative workflows.

AI Creative System Design · Prompt Architecture · JSON Schema Design · Workflow Development · Visual Direction
HTML / JavaScript · JSON · OpenAI · ComfyUI · Seedance
Functional prototype · R&D
Aurora began as an exploration of prompt engineering and creative education. A recurring challenge emerged: maintaining consistency between creative intention and AI-generated outputs. What began as a prompt framework evolved into a functional JSON-based prototype connecting creative strategy, AI direction and generative production workflows.
Turning creative intent into structured language.
Aurora transforms subjective creative decisions into structured inputs: brand voice, audience, visual rules, talent constraints and production requirements. The goal is not isolated images, but a reusable creative language.
A shared language between strategy and generation.
Aurora converts campaign inputs into a structured JSON object that carries creative decisions through different production stages.
{
"brand": {
"voice": ["direct", "playful"],
"visual_rules": ["product-first", "natural light"]
},
"campaign": {
"objective": "engagement",
"format": "ugc_vertical_video"
},
"production": {
"stages": ["concept", "image", "video", "review"],
"human_approval": true
}
}The schema is a contract between stages.
It defines brand voice, visual rules, audience, platform, identity constraints, adaptation rules and production targets while preserving human approval.
Structured direction reaches AI production.
The current prototype connects structured creative inputs with ComfyUI workflows, OpenAI-based creative direction and video generation. Testing focuses on product representation, character behavior, environment and visual language.

From prototype into creative infrastructure.
The next evolution explores automated storyboard generation, subtitle creation, format adaptation, creative scoring, campaign presets and multi-stage generation pipelines. These are future development areas, not currently implemented features.
Built, in progress and next.
The project clearly separates working functionality from systems still under evaluation.
Structured creative inputs, JSON generation contract, ComfyUI connection and AI-directed generation tests.
Consistency evaluation, brand fidelity and automated post-production.
Human-reviewed creative workflow infrastructure for repeatable campaign production.
- AI creative systems
- Prompt architecture
- Creative workflow design
- JSON-based production pipelines
- Generative video workflows
- Human-in-the-loop AI systems
- Creative automation
Unofficial R&D case study using Liquid Death as a test brand. No affiliation or endorsement is implied.