Aurora — AI Creative Production System
A functional R&D prototype that translates creative strategy into structured production instructions for human-reviewed image and video generation.
System overview — creative brief, structured JSON, generative production, campaign adaptations and human approval shown as one connected workflow.
AI Creative System Design · Prompt Architecture · JSON Schema Design · Workflow Development · Creative Direction
OpenAI · ComfyUI · JSON · HTML · JavaScript · Seedance
Functional R&D prototype
Aurora is a functional R&D prototype for turning a creative brief into structured, reusable production direction for AI-assisted image and video generation.
Instead of treating prompts as isolated instructions, the system carries the same brand, audience, platform and visual decisions through a normalized JSON contract, connected generation workflows and deliberate human review.
From creative intent to structured direction.
Creative campaigns often lose consistency as they move through different AI tools and production stages. Aurora addresses this by organizing creative decisions—including brand voice, audience, visual rules and production constraints—before generation begins. Rather than relying on disconnected prompts, every production stage references the same structured creative inputs.
One shared contract across the workflow.
Aurora converts approved campaign direction into a structured JSON object shared across the production workflow. The JSON acts as a shared contract between production stages while preserving human creative approval.
- Brand voice
- Audience
- Platform
- Visual language
- Production goals
- Review checkpoints
{
"schema_version": "aurora.master_input.v2",
"brand": {
"name": "Campaign brand",
"reference_notes": "Use approved assets"
},
"platform": {
"name": "TikTok",
"aspect_ratio": "9:16",
"duration_seconds": 15
},
"production": {
"batch_size": 2,
"shot_count": 4
}
}Structured inputs move through a working production chain.
The current prototype connects structured campaign inputs, AI interpretation, JSON generation, ComfyUI image workflows, video generation and human review.
- Visual consistency
- Product representation
- Character behavior
- Image-to-video continuity
- Prompt reliability
Evidence from internal R&D exercises.
Aurora has been tested through internal R&D exercises using publicly available brand references to evaluate structured creative workflows. Current experiments explore product campaigns, UGC-style video, image-to-video generation, creative consistency and brand interpretation.
Unofficial research exercise using Liquid Death as a fictional production scenario—from public brand references to product imagery and an image-to-video UGC test.
One reference image, structured shot instructions and an eight-second Seedance output.
The test evaluates character behavior, reference continuity, camera direction and time-coded scene changes. The result is shown as generated; human review remains the final checkpoint.
Working today. Evaluating now.
The current prototype separates demonstrated functionality from the qualities still being tested.
- Structured creative schema
- JSON generation
- OpenAI creative interpretation
- ComfyUI workflow integration
- Image generation
- Video generation
- Human review process
- Consistency
- Reliability
- Workflow optimization
- Brand fidelity
- Output quality
No roadmap. No hype. Only the current prototype and the evidence shown above.
- AI Creative Systems
- Workflow Design
- Prompt Architecture
- JSON Design
- Generative Production
- Human-in-the-loop Systems
Liquid Death is shown only as an unofficial research scenario using publicly available brand references. No affiliation or endorsement is implied.