ARC-StreamMemory: Building a Local-First Visual Second Brain for AI-Readable Video Memory

I’m building ARC-StreamMemory , a local-first visual memory system for AI-readable video, screen, snapshot, robotics, DAW/plugin, game, and app UI sessions.

The goal is to turn visual activity into something an AI can inspect, replay, cite, verify, and attach to a module.

Instead of treating video as a flat recording, ARC-StreamMemory turns it into a structured memory object:

visual source → FFmpeg video/snapshot ingest → AI frame-speed schedule → frame hashes → seeded source spine → OCR-ready/event-ready timeline → AI digest → ARC-style receipts → OmniBinary-style chunk map → Arc-RAR-style bundle manifest → local source-spine viewer → AI module attachment JSON

Enter fullscreen mode

Exit fullscreen mode

What ARC-StreamMemory does

ARC-StreamMemory can ingest visual sources such as:

video files

screen recordings

screenshots

DAW/plugin sessions

game footage

browser workflows

robotics camera feeds

app UI…