Website link to product walkthrough

The person brings a link to the product's site (and, best of all, the product to screenshot or a login to a demo account). You deliver a 45–90 s 16:9 walkthrough: a user's journey through the product on its own screens, step by step as the narration names each one, ending on the result and the brand's CTA. Contextual scenes can show when the product is needed if the brief benefits from them. Rounds and deliverables follow the core skill; record the flow and the picks where you keep the job's notes.

This is a product-demonstration recipe. The steps below own its production order; the website guide selects it without imposing another sequence.

Steps

1. Brand and facts (guide-capture-a-brand, route A; guide-pull-product-highlights, website route, plus the product on this computer when it runs here): the screens at full size, every step of the main flow in the product's own words. Gate: the flow to show, as a numbered list of 3–6 steps, and the screens for each.
2. Script (guide-make-a-script-from-source, a source to retell; the opener per llm-write-a-hook, the close per llm-write-an-ad-script): the moment of need, one beat per step in the flow's order, the change it made, the CTA in the site's words. About 2.7 words a second. Gate: the script word for word and the price.
3. Storyboard (hn4a-compositor-storyboard): the screenshots under the planned camera, a box for each generated moment, the bubble if there is a presenter. Gate: the direction, before any spend.
4. Voice (guide-voice-a-script, A narrator for a short), with a bed if wanted (elevenlabs-generate-music); cues on the take's own alignment (python-read-word-timings, cues). Gate: the voice.
5. The screens (guide-show-a-product-on-screen): choose each step's evidence. Use the actual recording to show an operation, a held screen for inspection, cut-up layers for a reveal, or a rebuild where data must change (hn4a-compositor-make-product-ui-exciting, hn4a-compositor-rebuild-product-screen). Camera pace follows tone and reading load (hn4a-compositor-fast-ad-camera, hn4a-compositor-calm-explainer-camera), with a hold where useful. Gate: each composed screen's test spec, played live.
6. Contextual moments, if chosen in the storyboard (guide-design-a-set, guide-animate-a-shot, guide-choose-a-video-model): show the situation around the product in the approved visual treatment. The demo used photoreal before/after scenes; these illustrate the need, not evidence of an actual event. Gate: the stills and a motion test when generating these shots.
7. Titles and the end (hn4a-compositor-brand-ad-titles): step numbers and names as each step starts, the logo and the CTA button on the end card; captions only if it will play muted (guide-caption-a-video); a presenter bubble only if the person wants a face (guide-present-with-an-avatar). Gate: the live spec (hn4a-compositor, Live spec, then render).
8. Deliver 16:9, and a 9:16 if it goes to social (guide-pace-a-short-for-social).

Choices this workflow makes

• The flow's order, not the homepage's: a walkthrough follows what a user does, screen after screen; the value props come in where a step shows them.
• Every step uses real product evidence. Show motion when it demonstrates the operation or directs attention; preserve time to read.
• Generated shots illustrate the context, if needed; keep the product UI real and choose the treatment from the brand direction.
• One screen per step at most: a step that needs two screens is two steps.

Dead ends

• None recorded on this workflow yet; the traps it inherits are in guide-show-a-product-on-screen and hn4a-compositor-make-product-ui-exciting. Record the job's own as it runs.

Worked example

The shape of the demo clip: Acme Desk, an invented after-hours phone desk for clinics.