voila

voila drives its own Chromium, records a product tour with an animated cursor and auto-zoom, narrates it with an open-source voice model on the CPU, and writes the MP4. It needs no screen-recording permission, no cloud and no API keys, and it embeds the script that rebuilt the video inside the file so anyone can fork or retarget the demo. Agents can plan, record, review their own frames and re-record through its MCP server.

How it works

1. 01

   Capture: a Chromium instance that voila owns streams frames over the DevTools Protocol at 2x pixel ratio while an animated cursor and every move and zoom are logged to a timeline.

2. 02

   Tour and narration: an auto tour or a YAML script supplies captions and narration, which the on-device voice model synthesizes first so each step lasts as long as its spoken clip.

3. 03

   Render: the timeline is replayed over the captured frames with eased zoom following the cursor and piped through ffmpeg into an H.264 MP4 with the recipe in its metadata.

What it does

• One command from URL to narrated MP4 with an automatic tour, or a YAML script for custom flows
• Narration and burned-in captions generated on-device, with a different voice or language per step
• The rebuild recipe is embedded in the MP4, so voila fork can retarget or recreate any voila demo
• MCP server with outline, record and review tools so an agent can watch its own video and fix it
• Desktop, mobile and tablet viewports; cookie banners are dismissed before recording, preferring reject