What Cassette.mp4 Is

Cassette.mp4 is a portable, self-describing context envelope for AI assistants. It packages transcripts, summaries, and instructions into a single file that any assistant can consume on drop. The format ensures repeatability, low drift, and portability by shrinking conversations or media into a predictable JSON schema that always plays back the same way.

• Condenses: Long transcripts and videos can be semantically chunked and summarized into compact payloads.
• Faithful: Using an 80/20 approach, the small cassette file can reproduce the original content close enough for accurate reasoning.
• Portable: Travels cleanly through email, chat, or uploads without corruption.
• Universal: Encapsulated in an MP4 container for compatibility across platforms.

Some mobile environments still treat novel hybrid payloads inconsistently; support will improve as adoption grows.

Why Cassette.mp4 Works

Most AI drift comes from feeding models raw, oversized transcripts or data. Cassette.mp4 reduces this risk by:

• Semantic chunking — slices transcripts or videos into meaning-rich segments.
• Summarization — creates a compressed but faithful retelling.
• JSON schema — enforces a predictable order and layout, so assistants always know how to read it.

This turns hours of dialogue or gigabytes of video into a slim, replayable context file.

File Anatomy

A typical project bundle includes:

• <ClientName>.mp4 — primary drag-and-drop cassette file
• cassette.json — same payload in JSON for editing/auditing
• schema.json / schema.min.json — optional JSON-LD for SEO
• script_tag.txt — ready-to-paste <script type="application/ld+json">…</script>
• README.txt — one-minute install + validation pointers

Minimum Payload

This illustrative skeleton shows the shape of a cassette without exposing orchestration logic. Field names are representative and may vary by project.

{
  "cassette_version": "1.0.0",
  "name": "...",
  "description": "...",
  "envelope": {
    "meta": { "...": "..." },
    "modules": [
      { "type": "summary", "ref": "kb://..." },
      { "type": "schema",  "ref": "schema://..." },
      { "type": "instructions", "ref": "prompt://..." }
    ]
  },
  "recipient": { "...": "..." },
  "signals": {
    "on_upload": "post:onboarding",
    "read_order": ["onboarding","context","install","faq"]
  },
  "content": {
    "placeholders": true,
    "notes": "Exact field names and execution details are provided to clients during implementation."
  }
}

Guided read order: onboarding → context payload → install steps → FAQ → appendix.

Memory Model (High-Level)

Cassette.mp4 supports a simple memory taxonomy to improve retrieval and repeatability:

• Episodic — references to past interactions/actions contained in the cassette
• Semantic — stable knowledge extracted or linked (e.g., summaries, schemas)
• Procedural — prompts, tools, and guardrails that shape assistant behavior
• Working — per-task scratchpad for what matters “right now”

These are references to existing or linked content—not duplicated data.

Example Uses

• Transcription archives: Meeting recordings distilled into portable cassettes.
• Video condensation: Long-form video compressed into semantic summaries and structured JSON (80/20 fidelity).
• Business onboarding: Schema delivery for client websites (e.g., JW Photographics).
• Knowledge bundles: Playlists, tutorials, or thematic libraries packaged for consistent assistant recall.

The business case is just one example. The broader innovation is condensed, drift-resistant context transfer.

Contact

Olsen Automation — Cassette.mp4 format inquiries, brian@olsenautomation.com

Developer Notes (Optional)

This page intentionally abstracts packaging and extraction specifics to preserve IP while communicating the concept and benefits. For pilots or custom integrations, contact Olsen Automation.