Make a product demo

This is the workflow we used for the 0.3 launch video.

1. Write the script. One Markdown file, with one short idea per scene. Scene length follows the voice, or the word count until you have one.
2. Record the product. Capture only what each scene needs, with clean test data. A recording of a few seconds per scene is enough.
3. Import the files in Assets.
4. Write headlines. Give each scene a title and a one-line subtitle in Storyboard. Headlines carry the story, so the capture can stay small.
5. Frame it. Mark a region or two, then move the camera to what the line is about. Start wide, and return wide.
6. Decide on AI, scene by scene. Use the real capture when it shows the point. Use an animation for ideas no screen shows.
7. Validate and approve. Fix what's listed, review privacy, approve.
8. Render, then watch it through with sound.

What you pay for

Real captures use no AI provider. Rendering runs on your computer with the Chrome and ffmpeg you already have. Costs come only from services you choose:

Step	Without a provider	With a provider
Captures, regions, camera	Included	Not needed
Rendering and re-rendering	Included	Not needed
Narration	Render silently with music	Your voice service
AI scenes	Paste a reply from a chat product	Your AI key

Generated results are stored, so once a scene is approved, later renders call no provider.

Tips

• Fix the recording, not the video. If a screen shows the wrong thing, record it again.
• Keep zoom modest. Record at a higher resolution than your video if you want to zoom. Validation blocks visible upscaling.
• Speed up before importing if a take runs long. Qontent Tube plays recordings at their own speed.