No edit points
Paragraph-sized ideas did not map cleanly to individual visual moments.
PRODUCTION SYSTEM · BUILT + DELIVERED
Turning a script, generated narration, timing data, motion components, and visual review into one controlled production workflow.
01 / THE FRICTION
A narrated explainer crosses several different systems: writing, voice generation, timing, visual design, component construction, animation, and rendering. A small change to the narration can move every beat that follows it.
If scenes are timed by feel or hardcoded directly into components, the production slowly becomes a collection of local fixes. Each correction creates another place for picture and sound to drift apart.
The narration needed to become production data—not just an audio file laid underneath the edit.
02 / BEFORE THE BUILD
The demonstration was built in stages, with rendered frames used as an audit trail. Four recurring failure modes shaped the finished workflow.
Paragraph-sized ideas did not map cleanly to individual visual moments.
Durations based on the script were always slightly different from the spoken delivery.
Timing placed inside components made every narration revision a cascade of edits.
A mathematically correct animation could still feel late, rushed, clipped, or simply wrong.
03 / THE SYSTEM
The script is divided into atomic beats. Narration returns character-level timestamps. Those timestamps become a scene schedule that every component reads. Nothing downstream gets to invent its own clock.
Break narration into short beats, each carrying one meaning and one visual idea.
Generate the voice and retain the timestamp attached to every spoken character.
Map each beat to exact visual start, speech start, speech end, and exit frames.
Drive reusable scenes and motion components from the shared schedule.
Render frames at sync points, watch the film, adjust, and render again.
THE OUTPUT
The finished film is the proof artifact. Its script, narration, scene timing, visual assets, motion, and rendered review all passed through the workflow described here.
04 / THE GUARDRAILS
No guessed duration. Audio timing comes from the generated narration, not estimated reading speed.
No overloaded scene. One beat carries one visual idea. A second idea requires a second beat.
No abrupt arrival. The visual begins before the spoken line and remains long enough for the moment to breathe.
No hardcoded clock. Components read their frames from the shared scene schedule.
No trusting the math alone. Every important sync point is rendered and inspected by eye.
05 / WHAT CHANGED
Estimate the narration length
Place scenes by feel
Store timing inside components
Fix drift one scene at a time
Discover problems in the final render
Generate narration with timestamps
Map speech directly to visual beats
Render from one scene schedule
Reuse a tested motion vocabulary
Inspect the exact frames that matter
THE PRINCIPLE
06 / THE HUMAN PART
Approval gates sit between the script, art direction, narration, timing, asset plan, and component architecture. The workflow preserves those decisions instead of silently carrying on. Software handles continuity and production mechanics; a person decides what the viewer should understand and how the moment should feel.
This case study is based only on the independently created Apollo 11 demonstration. It contains no client or employer material.
THE OUTPUT DOESN’T HAVE TO BE VIDEO
When several tools and people have to stay aligned, the most useful build may be the system between them.