Skip to content
All work
P-12activeStreaming Systems · 2026

StreamLab SXA — Low-Latency Secured Video Streaming

A low-latency, Tailscale-bound video streaming stack built around MJPEG capture, WebSocket metrics, adaptive buffering, and a browser dashboard for live telemetry.

Role
Developer · Streaming Stack & Telemetry UI
Timeframe
May 2026
Domain
Streaming Systems
Status
active
Peak FPS
25 FPS
Lowest Latency
3.15 ms
Peak Bitrate
11.27 Mbps
StreamLab SXA — Low-Latency Secured Video Streaming

Overview

StreamLab SXA is a camera streaming system that separates capture, transport, client adaptation, and dashboard telemetry into a compact local network workflow.

The stack is designed so the server emits MJPEG frames, the client measures stream health in real time, and the browser dashboard exposes the operational state without needing a heavyweight backend.

The Challenge

The real problem was not sending video. It was keeping latency visible, frame loss bounded, and the whole pipeline reachable only through a controlled network path instead of exposing it like a fragile public toy.

Approach

  • 01Implemented server-side capture and MJPEG delivery through `server.py`.
  • 02Pushed live operational metrics over WebSocket so the dashboard could reflect stream health in real time.
  • 03Built the client around adaptive buffering, frame-gap detection, and loss estimation.
  • 04Kept the services bound to Tailscale-only access to avoid broad network exposure.

Outcomes

  • Delivered a measurable low-latency streaming pipeline with client and server telemetry.
  • Added practical loss visibility through FEC markers, buffer tuning, and logging.
  • Provided a browser dashboard for live monitoring and remote operational control.

Stack

Backend

  • Python
  • aiohttp
  • websockets
  • OpenCV

Client

  • Python
  • requests
  • NumPy
  • Adaptive buffer

Interface

  • HTML
  • CSS
  • JavaScript
  • Browser dashboard

Interested in a build like this?

I take on instrumentation, embedded, and computational physics work. Let's talk about your measurement problem.

Gilang Pratama Putra Siswanto