---
title: "VisionPiper Docs — ModelPiper"
description: "Documentation for VisionPiper. Real-time camera and screen analysis with local vision models. OCR pipelines and custom vision workflows."
canonical: "https://modelpiper.com/docs/visionpiper/"
---

# VisionPiper Docs — ModelPiper

> Documentation for VisionPiper. Real-time camera and screen analysis with local vision models. OCR pipelines and custom vision workflows.

# VisionPiper

Offline

Standalone macOS menu bar app for screen capture, recording, GIF conversion, and live WebSocket streaming.

## Capabilities

### Region Capture

Select any screen region with an 8-window border architecture (4 edges + 4 corners) and pass-through interior.

### Recording

SCStream + AVAssetWriter for H.264 .mp4 output. GIF conversion via vendored cgif + libimagequant.

### Live Streaming

WebSocket on port 10000. Metal JPEG frames streamed to browser via `createImageBitmap()`.

## Build & Run

`./build.sh --visionpiper` — incremental build, preserves TCC permissions. Add `--clean` to reset DerivedData, `--reset-tcc` to re-grant screen recording access.
