Launch video

The person brings what is launching (a changelog entry, a release note, the site, the product itself) and who should present it. You deliver a 30–90 s launch video: the presenter talking to camera beside the product's real screens, the brand's kinetic titles landing on the key words, karaoke captions, a name strap and the CTA, in 16:9 and 9:16. Rounds and deliverables follow the core skill; record the answers and picks where you keep the job's notes.

Steps

1. What's new (guide-pull-product-highlights: the changelog, release notes and site, or the product running here for screenshots) and the brand (guide-capture-a-brand). The one thing that is new, in the product's own words, and the one thing the viewer should do.

2. Script in the launch arc (llm-write-an-ad-script; the hook per llm-write-a-hook; the length per guide-make-a-script-from-source): the problem → "today we're launching …" → what's new, a beat per change → proof (a real number, a user) → the CTA. Written to be said, for the presenter's own voice. Gate: the script word for word.

3. Ask: "Will you record a quick webcam video reading it, or should an AI avatar of you present it?" Then follow the answer:

   They pick	Do
   A webcam video	They record the approved script (front light, a still camera, a few takes of each line are fine); cut the best takes (guide-cut-best-takes, route A) and transcribe the cut (whisper-cli-transcribe-with-word-times). Gate: the rough cut
   An AI avatar	guide-present-with-an-avatar, the real-person route: the still from their own photos, their cloned voice with consent or a picked one, the whole track lip-synced after step 4. Its gates stand

4. Storyboard (hn4a-compositor-storyboard): the presenter (the cut, or a silhouette) at their size and side, the screens, the title stack, the strap and the captions box. Gate: the direction, before any lip sync or rebuild.

5. The product on screen (guide-show-a-product-on-screen): the new thing's own screen or recording, held for reading or revealed when that clarifies the change (hn4a-compositor-make-product-ui-exciting), or a card beside the presenter.

6. Titles, strap and captions (hn4a-compositor-brand-ad-titles): the hook stack, the launch's name slammed in, chips for what it works with, a sticker for a price or "free", the end card with the CTA; a name strap with the presenter's name and role (hn4a-compositor-chyron, the name strap); karaoke captions in the brand (guide-caption-a-video). A bed and a few hits if wanted (elevenlabs-generate-music, guide-add-sound-effects). Gate: the live spec (hn4a-compositor, Live spec, then render), both aspects.

7. Deliver 16:9 and 9:16, −14 LUFS for the 9:16 posted to social (guide-pace-a-short-for-social).

Choices this workflow makes

• An AI avatar is labelled as one: the strap reads "AI ", so nobody watching takes it for a recording.
• The webcam route is free and the most real; the avatar is for a founder who won't record, or for a launch re-cut into many versions.
• The presenter at one side, the product and titles at the other: the screen and type get half the frame, the person stays on screen the whole time.
• For more motion over a real recording (titles that react to their hands, type behind them), carry on in workflow-add-captions-and-motion-graphics-to-video on the cut.

Dead ends

• None recorded on this workflow yet; the traps it inherits are in guide-present-with-an-avatar and guide-cut-best-takes. Record the job's own as it runs.

Worked example

The shape of the demo clip: an AI avatar of the founder, from her own photos and voice, on the right of a dark frame with a faint green glow; on the left a terminal card types the product's one command, a "FREE" sticker pops beside it, and a two-line title stack slams in on the words; a "Works with" row of chips follows; karaoke captions run along the bottom over a name strap reading "AI , CEO".