Complex spoken explainer
A 137-second Chinese explainer built from 17 accepted performance clips, recorded demonstrations, project-owned MG, independent Caption and sound. The example demonstrates useful component boundaries without turning every design constant into a public parameter.
Start here
From this production directory:
hypit check authors/main.svml
hypit plan authors/main.svml --target complete-film.video
hypit studio --source authors/main.svml --target complete-film.video
The accepted performances and supporting media are ordinary project assets selected by the Author Source. Read the plan before submitting a Build; speech evidence is still produced by the explicitly authored transcription declarations unless the Source is changed to use project-owned transcript evidence.
Where an edit belongs
| Edit | Owner |
|---|---|
| Spoken words, Cue boundaries, Moments and Selections | authors/script.svml |
| Input images, recordings, fonts and clock/canvas | authors/assets.svml |
| Performance prompt and selected speaker Kit | authors/direction.svml, recipes/performance.svs |
| Material outputs, media placement, Uses, scene events and Film | authors/main.svml |
| Caption appearance and ordinary presentation | recipes/composition.svs |
| A coordinated visual behavior | The owning project package under ../../packages/ |
| Which accepted performance file is used | authors/assets.svml and authors/main.svml |
Generated-material and preparation relations remain explicit in the entry Source. Imported Sources own Script, raw resources and direction; the current command selects Targets.
Components
opening-system owns opening graphics, flags, stage and presenter framing Styles. web-scenes owns eight coordinated demonstrations. single-line-captions presents Script cues through Caption’s semantics and public Companion interfaces. launch-scenes supplies the montage poster. visual-language is an ordinary shared palette/helper dependency. Each package README explains its inputs and ownership.
Export
Select complete-film.video to render the complete film. For machines with limited temporary disk,
select export-part-1.video through export-part-5.video in one command by repeating --target.
Each Build is a fresh execution attempt. The local profile currently requests eight rendering workers.
After exporting the five named Outputs with hypit get, join their video streams without re-encoding:
python3 scripts/assemble-final-export.py part-1.mp4 part-2.mp4 part-3.mp4 part-4.mp4 part-5.mp4 --output final.mp4
The helper derives each part’s length from its encoded video, removes each AAC part’s encoder delay before joining sound, and encodes one final audio stream. Parts must be consecutive, ordered and made with the same video settings.
Read Treatment, craft notes and asset provenance. Private exploration and old outputs are preserved in history, outside the portable example.