I wanted to share a slightly unusual Dash project: instead of a dashboard for charts, I built a full custom OBS-style live streaming system with it. It drives multiple Insta360 cameras, lets me switch/zoom viewports from a physical Stream Deck, and includes an AI “director” that tracks movement and re-frames the scene. State lives in Postgres and the app streams out to YouTube/Twitch/Facebook.
The most useful practical insight for anyone building media-heavy Dash apps: do not try to shuttle images through your app as base64 — offload all media to a CDN (I use Cloudflare) so the Dash app only handles control/logic, not the heavy payloads. That single decision is what made the real-time viewport switching feel instant. Dash 4.2 performance improvements also made a big difference here. I also keep each feature as a documented, self-contained component with an llms.txt so I can rebuild and extend fast.
Full walkthrough here if you’re curious this is the first live stream I created directly from the 2plot.media app:
Dash docs/components I lean on:
https://2plot.dev and https://dash-mantine-components.com — and the platform this powers:
check out the ai canvas showcased in this video: