Your own script

What it is. You write the narration in a plain text file. The kit reads it out word for word. Nothing rewrites it.

Steps

1. Write script.txt beside the config. One paragraph per scene, in order, with a blank line between them:

   The shipment file came in again this morning.
   
   One click keeps only the shipments whose date moved.
   
   Tideline. Every change found, and kept.
   

2. To tie a paragraph to one scene, whatever its place in the file, put the scene's id in a heading above it:

   ## end
   Tideline. Every change found, and kept.
   

3. Point the config at it:

   { "url": "http://localhost:3000", "script": "script.txt" }
   

4. In video.ts, leave say out of the scenes. Each still needs an id:

   { id: 'hook', motion: ({ theme, sayMs }) => hookCard(theme, { title: 'The file. Again.' }, sayMs) },
   

5. Render:

   bunx showcase make --draft
   

Check it worked

• The captions in the video are your sentences, unchanged.
• If the number of paragraphs does not match the scenes, the run stops and names the scenes that have no line.
• If the script is more than 10 percent over the word budget for the length, you get a note and the video is still rendered:

  note: the script is 100 words; the chosen length fits about 70. Cut about 30 words, or choose a longer video.
  

Good to know

• A scene that already has say in video.ts keeps it; the script fills only the scenes without one.
• Write for the ear: short sentences, one idea each. Avoid exact counts and dates if the product's data changes.