Sightline · archived · open source

Real-time screen detection for stream automation.

A Windows desktop app and runtime that watches a live frame source, detects visual or OCR-based events, and runs follow-up actions: OBS scene and source changes, sounds, program launches, chapter markers, and autoclip workflows. It is no longer in active development; the source is public.

Source on GitHub Watch demo

Visual template and OCR detection for on-screen events
OBS actions, sounds, app launches, and chapter markers
Autoclip workflows driven by detections

Sightline live detection preview in Warzone

Status Archived · source available on GitHub

Profiles, detection previews, action chains, and live runtime status in one workflow.

Demo footage

Sightline running live.

Detection, trigger configuration, and highlight workflows running against live gameplay.

If the embed is blocked by browser privacy settings, watch it directly on YouTube.

Features

What it does.

Visual detection

Define screen regions, preview matches, and tune detection against the preview before a trigger goes live.

Action workflows

Trigger sounds, launch programs, update OBS text sources, switch scenes, or toggle scene items when an event fires.

Autoclipping

Feed successful detections into an autoclipping pool so moments are preserved during live sessions.

Workflow

From signal to reaction in three steps.

Capture the right region

Point Sightline at the on-screen area that matters and preview the detection target.

Pick the reaction

Choose what happens next, from sounds and programs to OBS scene and source changes.

Turn hits into highlights

Route successful triggers into autoclipping for a shorter path from stream to clip.

Sightline detecting a live gameplay event — 1. Detect the moment live in the scene.

Sightline previewing the capture region and detector result — 2. Preview the match before it goes live.

Sightline action workflow configuration — 3. Fire actions and route hits into clips.

Screenshots

Setup and live control.

Sightline live detection view during gameplay — Live detection running against gameplay with regions layered directly over the action.

Sightline preview region modal — Preview the crop, processed result, and detection confidence before committing a trigger.

Sightline actions workflow editor — Chain actions and advanced workflows from a single event.

Sightline profiles view — Per-game profiles, each setup isolated and reusable.

Under the hood

How Sightline is put together.

The core loop: watch a live frame source, detect a visual or OCR-driven event, run follow-up actions, and optionally route the hit into clipping workflows.

Perception core

perception_core is a reusable library for region geometry, image and template detection, OCR text extraction, preprocessing, and the schema that detection profiles are saved in.

Runtime

sightline_runtime handles capture providers (including OBS), event gating and emission policy, action workflows, OBS scene actions, chapter markers, and output sinks.

Desktop editor

A pygame-based desktop app for authoring profiles, previewing regions and OCR settings, and building trigger and action workflows, with a local OBS overlay service for on-stream feedback.

Packaging

Compiled with Nuitka and packaged with an Inno Setup installer for Windows. The build script still supports the free and premium tiers the app was originally distributed as.

Source & status

Archived, with the source public.

Sightline is no longer in active development. The codebase is on GitHub at SessionGoblin/sightline-computer-vision. Running it from source needs Windows, a separate Tesseract OCR install, and OBS with obs-websocket for live capture and OBS actions.

Source on GitHub More projects