Pillar — Local-First

Local-First Streaming. Sub-120ms. No Cloud Tax.

VPE's decision layer runs entirely on your PC. Scene switches in 12ms p50. Clip captures in 47ms. Chat actions in 89ms. No cloud round-trips, no browser-source lag, no third-party server holding your chat logs. The architectural pillar that makes everything else fast.

See How It Works

12ms scene switch p5047ms clip capture89ms chat actionZero cloud round-tripsYour data stays on your PC

Latency — Measured, Not MarketedLive

Twitch event arrives → VPE local pipeline

Cloud tool: ~340ms p50 (platform → cloud → browser source → OBS)

VPE: 12ms p50 (platform → local engine → OBS)

28× faster path. Same result. Zero external dependency.

12ms

Scene switch p50

47ms

Clip capture

89ms

Chat action

Cloud tool path (≈340ms p50)

Twitch event
  → cloud relay (NA East / 80ms)
  → cloud decision worker (40ms)
  → browser source poll (~150ms p50)
  → OBS scene change
  ─────────────────
  ≈ 340ms p50 end-to-end

VPE local path (12ms p50)

Twitch event
  → local VPE engine (decision in 4ms p50)
  → local obs-websocket (127.0.0.1, 8ms p50)
  → OBS scene change
  ─────────────────
  ≈ 12ms p50 end-to-end

Sub-120ms p50 end-to-end across every action

Your chat logs never leave your PC

Your donation data never leaves your PC

Engine survives internet drops, reconnects cleanly

No third-party SaaS lock-in or data egress fee

Audit-friendly for agencies handling client streams

Works with the same OBS, scenes, and overlays you have

Zero cloud usage fees — free tier costs you nothing

How It Works

The Architecture That Makes Local Possible

Direct Platform Websocket

VPE connects directly to Twitch EventSub, YouTube Live Streaming API, and Kick's Pusher channel. No middleman cloud relay. Events arrive on your PC the moment the platform fires them.

Local Decision Pipeline

The 6-layer semantic pipeline classifies each event, scores it, and chooses a response — all in process, in your RAM. No HTTP call, no queue, no scheduler in a foreign datacenter. The decision exists before any cloud would have even answered the ping.

Local OBS Control

OBS Studio runs on the same PC. VPE talks to it over the local obs-websocket on 127.0.0.1. Scene switches, source toggles, and alert overlays fire over a localhost socket — which is as fast as IPC gets on consumer hardware.

Stop paying the cloud tax. Run automation that's actually fast.

← Home

← Multistream For Agencies →

Frequently Asked

What does local-first stream automation mean?

Local-first means the decision layer of your stream automation runs on your PC — not in a third-party cloud. Platform events arrive directly over the platform's own websocket, VPE classifies them locally, and the response (scene switch, alert, clip, chat action) is dispatched without a round trip through an outside server.

Why is sub-120ms latency important for streaming?

Cloud-based stream tools commonly add 300–500ms of round-trip latency before an alert fires or a scene switches. That's enough to make raids land after the energy has passed and to make donation alerts feel awkward. Sub-120ms means the response is faster than a viewer's reaction time — the alert fires while the moment is still happening.

Does local-first mean offline-only?

No. VPE still talks to streaming platforms (you need internet to stream). The difference is that the automation layer itself runs locally, so if your home internet hiccups, the engine keeps running and reconnects cleanly when the platforms come back. Cloud tools die entirely until their servers reach you again.

What data leaves my PC with VPE?

Stream traffic goes to the platforms you're broadcasting to (Twitch, YouTube, Kick, etc.) — there's no way around that. Beyond that, telemetry is opt-in, chat logs never leave your machine, donation events never leave your machine, and clip files stay on your drive until you choose to publish.