All products

Compose a talking-head video with AI

Combine prepared landscape footage, subject masking, timed text and premixed audio into a reviewed talking-head export.

Compose a talking-head video with AI from prepared picture, subject masking, text timing and sound. The choosing-video skill applies a landscape preset with white headings behind the speaker, pale-yellow bottom captions and floating picture cards.

Its renderer combines those inputs; it does not automatically transcribe, segment the subject, choose takes or mix audio. Keep every layer aligned after timeline changes. The public package includes the title font and settings, while personal reference footage, the retained project and its exact soundtrack remain outside the package.

What it does

  • Place large headings behind a moving speaker
  • Reveal timed captions at stable phrase anchors
  • Keep the subject matte aligned with picture
  • Add prepared floating cards with muted audio
  • Use a complete premixed soundtrack
  • Review previews before native-source export

Install and use

  1. Load the complete folder in a file/shell-capable agent. Prepare Python 3.10+, Pillow and imageio-ffmpeg, plus licensed local subtitle typography or an explicitly selected licensed replacement. The title font is included.
  2. Supply an edited 16:9 assembly, aligned matte, word events, prepared cards and final audio. Ask: “Combine this teaching clip with Latin headings behind the speaker and two prepared cards. Preserve framing and the spoken ending.”
  3. Inspect a preview for readable titles, hair and hand occlusion, card identity and speech alignment. Any cut requires retiming all dependent layers together.
  4. Watch and listen to the complete result. Full decoding checks structure, not perceptual quality. Larger exports retain real detail only when prepared layers come from native-resolution sources.

FAQ

Source

Made by bhavya-gx. Original skill: choosing-video, published on GitHub under the MIT licence.

View on GitHub