Overview
Shipping a feature usually ends with someone screen-recording a shaky walkthrough. Roomote can produce that video for you: it drives the product in its sandbox browser, records a clean clip, and adds a synthetic cursor, click ripples, and captions. With narration configured, the demo gets a voice-over that leads each visual beat.- Trigger: Ask in a task
- Setup time: None for captioned demos; 5 minutes for narration
- Requires: Nothing extra; admin access for narration
- Serves: Engineers, Product, Marketing
Ingredients
- Any task where the feature is reachable: a running dev environment or a public page
- Optional: an ElevenLabs API key and voice ID for narration
Steps
- In a task, ask for a demo, for example:
Make a demo video of the new custom MCP servers settings pageor invoke$feature-demodirectly. - Roomote plans the story beats, records the interaction live, and renders a 16:9 video (ask for a vertical cut for short-form if you want one).
- The finished video is uploaded as a task artifact and linked in the reply
(and embedded under
## Screencastswhen the task opens a PR).
Narration
Captioned demos work out of the box. For a voice-over in your own (cloned) voice, a deployment admin connects ElevenLabs under Settings → Integrations → ElevenLabs with:- an ElevenLabs API key, scoped to text-to-speech only with a credit cap (sufficient and recommended), and
- the voice ID to narrate with.
R_ELEVENLABS_API_KEY and R_ELEVENLABS_VOICE_ID environment variables;
the integration takes precedence when both are present.)
Tips
- Narration lines come from the on-screen captions; ask for wording changes the way you would edit copy (“make the second line more casual”).
- The narrative drives the pacing: each shot holds for as long as its line takes to speak, voiced or not.
- Keep demos to 2-5 beats. Two shorter videos beat one long one.