ARC-StreamMemory

May 6, 2026 · View on GitHub

Local-first visual second brain for AI-readable video, screen, snapshot, robotics, and source-spine memory.

Repo Version Python FFmpeg Local First

ARC-StreamMemory turns video files, screen recordings, screenshots, DAW/plugin sessions, game footage, browser work, robotics camera feeds, and app UI states into deterministic, cryptographically hashed, AI-readable visual memory modules.

It is designed as the visual second brain / AI sight spine for the GareBear99 ARC ecosystem: frames become indexed evidence, summaries become module attachments, hashes become source-spine proof, and capture sessions become replayable memory bundles.


Quick answer

ARC-StreamMemory is for anyone searching for:

  • AI visual memory
  • visual second brain
  • video memory for LLMs
  • screen recording for AI
  • local-first multimodal memory
  • visual RAG / frame retrieval
  • deterministic video archive
  • cryptographic frame hashing
  • robotics camera memory
  • FFmpeg frame sampling for AI
  • reproducible visual evidence bundles
  • AI-readable screenshots and session replay

What it does

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

The output is not just an MP4 or screenshots. It is a structured memory object that another AI can inspect, replay, cite, and attach to a module.


Why this matters

Normal screen recording answers:

What happened? Maybe watch the whole video again.

ARC-StreamMemory answers:

What happened?
→ Read the AI digest.
→ Jump to event 000042.
→ Open frame_000042.
→ Verify the frame hash.
→ Follow the ARC receipt.
→ Follow the OmniBinary-style pointer.
→ Package/restore through the Arc-RAR-style bundle manifest.

That makes the memory AI-readable, deterministic, replayable, and source-verifiable.


Current status

CapabilityStatusNotes
Demo generated visual-memory session✅ completedeterministic proof session
Snapshot folder ingest✅ completescreenshots/images into AI memory
Regular FFmpeg video ingest✅ complete if FFmpeg is installedsample MP4/MOV/MKV/WebM at chosen FPS
AI frame-speed policy✅ complete0.2, 0.5, 1, 2, 5, 10 FPS-style workflows
Frame hashing✅ completeSHA-256 per frame
Memory spine hashing✅ completedeterministic session root lineage
Seeded source spine✅ completereplayable seed/hash lineage
AI digest✅ completeMarkdown + JSON digest
AI module attachment✅ completestructured object for AI consumption
ARC-Core-style receipts✅ completeexport format ready
OmniBinary-style chunk map✅ completepointer format ready
Arc-RAR-style bundle manifest✅ completeZIP export today, Arc-RAR integration later
Local HTML viewer✅ completeinspect frames, events, digest, JSON
ARC-FusionCapture adapter layer✅ complete as adapter/specrobotics-ready capture policy + sensor sync scaffolding
Native live screen capture🚧 integration gatenot falsely claimed yet
Real OCR engine🚧 integration gateplaceholder/index format ready
Native OmniBinary persistence🚧 integration gateexport map ready
Native Arc-RAR packaging🚧 integration gatemanifest ready

Quickstart

python scripts/create_demo_session.py
python scripts/build_stream_memory.py examples/demo_session --title "ARC demo visual memory"
python scripts/hash_memory_spine.py examples/demo_session
python scripts/build_seed_spine.py examples/demo_session
python scripts/build_ai_digest.py examples/demo_session
python scripts/validate_memory_bundle.py examples/demo_session
python scripts/make_bundle.py examples/demo_session --out release_evidence/demo_streammemory_bundle.zip

Open viewer:

python -m http.server 8080

Then open:

http://localhost:8080/viewer/index.html?session=../examples/demo_session/memory/module_attachment.json

Regular FFmpeg workflow

ARC-StreamMemory works with normal FFmpeg today.

python scripts/ffmpeg_probe.py
python scripts/ingest_video.py input.mp4 --fps 1 --out sessions/video_memory
python scripts/build_stream_memory.py sessions/video_memory --title "Video memory"
python scripts/hash_memory_spine.py sessions/video_memory
python scripts/build_seed_spine.py sessions/video_memory
python scripts/build_ai_digest.py sessions/video_memory
python scripts/validate_memory_bundle.py sessions/video_memory

Recommended AI frame rates:

FPSUse case
0.2long passive session memory
0.5lightweight visual diary
1general AI inspection default
2UI debugging / GitHub / DAW flows
5detailed interaction review
10motion-sensitive review

ARC-FusionCapture compatibility

ARC-StreamMemory now includes a compatibility layer for the planned ARC-FusionCapture runtime.

ARC-FusionCapture is the future robotics/media capture layer that should wrap regular FFmpeg with:

  • camera/feed profiles
  • robotics capture modes
  • hardware-acceleration selection
  • sensor timestamp sync
  • rolling buffer policy
  • event-triggered clips
  • AI-friendly frame-speed output
  • ARC-Core receipts
  • OmniBinary pointers
  • Arc-RAR bundle manifests

Current files:

integrations/arc_fusion_capture/README.md
integrations/arc_fusion_capture/profile_presets.json
scripts/build_capture_policy.py
scripts/fusion_capture_adapter.py
scripts/sync_sensor_log.py
schemas/capture_policy.schema.json
schemas/sensor_frame.schema.json
schemas/robotics_session.schema.json

Example robotics-style adapter flow:

python scripts/build_capture_policy.py --mode robot_navigation --fps 2 --out sessions/nav_test/memory/capture_policy.json
python scripts/fusion_capture_adapter.py --input robot_run.mp4 --policy sessions/nav_test/memory/capture_policy.json --out sessions/nav_test
python scripts/sync_sensor_log.py sessions/nav_test examples/sensor_log.jsonl
python scripts/build_stream_memory.py sessions/nav_test --title "Robot navigation memory"
python scripts/build_seed_spine.py sessions/nav_test
python scripts/validate_memory_bundle.py sessions/nav_test

Output structure

session/
├─ frames/
├─ memory/
│  ├─ capture_policy.json
│  ├─ frame_index.json
│  ├─ event_timeline.jsonl
│  ├─ ocr_index.jsonl
│  ├─ ai_digest.md
│  ├─ ai_digest.json
│  ├─ module_attachment.json
│  ├─ memory_spine.json
│  ├─ seed_spine.json
│  └─ session_summary.md
├─ receipts/arc_receipts.jsonl
├─ omnibinary/chunk_map.json
├─ arcrar/bundle_manifest.json
├─ reports/validation_report.json
└─ reports/bundle_export_report.json

Seeded source-spine model

ARC-StreamMemory stores visual memory with a deterministic seed chain:

capture_policy_hash
+ source_fingerprint
+ frame_schedule_hash
+ ordered_frame_hashes
+ chunk_hash
= session_root_seed

This gives the session a reproducible source spine:

root_seed
→ chunk_000001
→ frame_000001
→ frame_hash
→ event_receipt
→ module_attachment_pointer

This pattern was strengthened using the seeded replay/source-lineage direction from Seeded-Universe-Recreation-Engine.


Ecosystem repo references

ARC-StreamMemory is part of a linked ARC / TizWildin / GareBear99 ecosystem. Every repo below is either a source reference, benchmark target, integration target, discovery surface, or public-router anchor.

RepoRole in ARC-StreamMemory
ARC-Coreauthority layer, receipts, event truth, source governance
omnibinary-runtimebinary-addressable memory spine and chunk-ledger inspiration
Arc-RARportable archive/restore bundle direction
ARC-Neuron-LLMBuilderlocal AI memory, governed build loop, module attachment use case
ARC-AudioBenchbenchmark/evidence methodology companion
Seeded-Universe-Recreation-Engineseeded source-spine / deterministic replay storage model
FreeEQ8DAW/plugin testing screen-memory target
FreeVox8spectral plugin / visual-memory benchmark target
Voxel Audioaudio-visual router/session target
InstrudioWeb Audio / MIDI / virtual instrument memory target
awesome-audio-plugins-devStart Here technical discovery list
awesome-audio-listsroot hub for audio/list/discovery resources
awesome-python-audio-scienceresearch/citation bridge for Python audio science, MIR, DSP, ML audio
awesome-music-platformsartist/music platform discovery map
TizWildinEntertainmentHUBpublic router for plugins, lists, packs, deconstructed loops, visualizers
TizWildin-Release-Vaultrelease surface for music assets, deconstructed loops, and packs

Public-facing use cases

AI developer

Use ARC-StreamMemory to turn debugging videos and UI sessions into reproducible visual memory modules.

Audio/plugin developer

Use it to archive DAW/plugin tests, FreeEQ8/FreeVox8 validation sessions, pluginval videos, and UI regressions.

Robotics developer

Use regular FFmpeg now, then connect ARC-FusionCapture later for sensor-synced camera memory and robot black-box replay.

Research / reproducibility

Use CITATION.cff, seeded spines, hashes, and module attachments to make visual sessions citable and reproducible.

Creator / game developer

Use it to capture game states, UI flows, visual bugs, and build history as replayable evidence bundles.


ai-memory
visual-memory
video-memory
screen-recording
multimodal-ai
visual-rag
local-first
ffmpeg
robotics
computer-vision
cryptographic-hashing
reproducible-research
second-brain
arc-core
omnibinary
arc-rar

Honest completion status

This package is complete as a public source release foundation for deterministic visual memory:

  • ingest
  • indexing
  • hashing
  • seeded source-spine storage
  • AI digest
  • module attachment
  • viewer
  • validation
  • regular FFmpeg workflow
  • ARC-FusionCapture adapter layer
  • bundle export

Remaining integration gates are intentionally not overclaimed:

  • native live screen capture
  • real OCR engine hookup
  • native OmniBinary persistence
  • native Arc-RAR packaging
  • live ARC-Core API sync
  • production robotics sensor bus integration

License

MIT License.