Moon
A longer science storytelling experiment exploring the Moon’s place in our world. It brings narrative, scene production, visual explanations and sound into a complete film.
We are developing an end-to-end AI-assisted workflow that connects research, scripting, production planning, procedural visuals, rendering, independent review and quality control. Built for our own channels and client media projects, with creative direction and verifiable outputs at every stage.
Founded in March 2022, antiblue studio brings hands-on experience in client media work and YouTube production. That experience shapes the problems we are solving: fragmented tools, repetitive production tasks and fragile handoffs between planning, execution and review.
The goal is a connected production workflow, not a collection of isolated AI outputs. Specialized agents and production tools work from structured plans, pass artifacts between stages and preserve the state needed to inspect, revise and resume the work.
Useful media needs substance, a clear narrative and purposeful visuals. We combine editorial and technical review with programmatic checks, so a successful render is only one step toward an accepted output. Human direction remains part of the process.
Two outputs from our production research. Science films are a testbed for the broader system we are building for owned channels and client media.
A longer science storytelling experiment exploring the Moon’s place in our world. It brings narrative, scene production, visual explanations and sound into a complete film.
A short, code-driven exploration of water and phase change. Developed with Claude, this experiment connects creative direction with procedural scenes, animation, rendering and audio.
Science films are our current testbed for the wider production system. These experiments connect source research and narrative planning with procedural visuals, rendering and independent review of both scientific meaning and presentation.
Code-driven scenes make it possible to revise geometry, motion, lighting and camera choices at source level. Experiments across Blender, Godot, Three.js and Babylon.js explore how reproducible visual production fits into a larger media workflow.
An execution layer coordinates bounded tasks, tracks dependencies and checks the artifacts returned by each worker. Our experiments explore persistent state, restart and recovery, failure handling and independent acceptance: completion claims alone do not establish that the work is done.
Each stage produces inputs for the next. Review can send work back for a targeted revision; delivery follows acceptance, not just generation. The workflow is being developed and tested through real production experiments.
Gather sources, examine claims and organize the research that shapes the story.
Turn grounded research into a script, scene plan, timing, asset requirements and production specifications.
Create and stage procedural scenes, visual assets, narration and audio against the production plan.
Render scenes and compose picture, audio and captions into inspectable outputs.
Review scientific meaning, storytelling, visual clarity and technical output. Feed identified issues back into revision.
Package the reviewed media, source references and production records for channel or client delivery.
Our answer to low-quality AI content is a production process that asks what the work says, why each visual is there and whether the output holds up to review.
Research and source checks inform the script. Scientific claims and visual explanations are reviewed for meaning, not just fluency.
Diagnostic renders and independent review help examine composition, motion, pacing and technical defects. Issues lead to source revisions and another check.
Structured plans, checkpoints and artifact records connect the intended work with the actual result. Verification and human judgment guide acceptance.
We connect reasoning models with procedural engines and media tools. Each tool has a defined role; structured handoffs and verification make the workflow more important than any single model.
We have used Claude for scripting, planning, tool debugging and creative experiments in 3D video. Our current workflows combine multiple AI tools; direct Claude API integration is a next step to evaluate for selected reasoning, coding and review tasks.
Real-time rendering runtimes used for procedural background generation, interactive camera staging, and programmatic graphics synthesis without conventional manual UI bottlenecks.
Headless Python API integration for procedural geometry creation, material graph generation, automated lighting passes, and batch rendering workflows.
Complex filtergraphs for real-time video compositing, SMPTE-aligned subtitle styling, audio ducking, LUFS loudness measurement and normalization, and automated defect auditing.
From hands-on media production to a connected AI-assisted workflow.
antiblue studio is a self-funded media company founded in March 2022. Its work spans client media production and operating its own channels. Today, founder Nguyen Tuan Duc leads the development of an internal production system informed by that practical experience.
The system is designed to support channel content and client projects, rather than to sell the underlying tools as a software product. Science films are the current testing ground. Our focus is connecting the entire workflow while preserving editorial judgment, creative intent and checks on the finished work.
For technical discussions, research exchange, or company verification inquiries regarding antiblue studio:
Official communications domain: antibluestudio.com