Transcribe Microsoft Teams on Mac without bots

Microsoft Teams stands as the enterprise communication backbone for global financial institutions, healthcare networks, defense contractors, and technology leaders. On corporate Mac laptops, professionals spend hours each day navigating architecture reviews, cross-functional committee sessions, and incident triage channels within Teams.

Capturing accurate notes from these discussions has historically forced a painful choice: submit to Microsoft’s exorbitant Copilot licensing fees or attempt to use third-party cloud bots that enterprise information security teams routinely ban.

Oyma resolves this friction through desktop isolation. Operating directly within macOS, Oyma captures and transcribes Teams conferences locally on Apple Silicon without requesting tenant administrative consent, without routing audio to third-party clouds, and without deploying an unapproved bot into the call lobby.

The enterprise bottleneck: Tenant bans and the Copilot premium

In enterprise Microsoft 365 environments, capturing call records using typical market tools hits formidable organizational and architectural walls:

1. Strict tenant bot restrictions: Enterprise M365 administrators routinely disable external guest participation and block automated recording applications. When a cloud service attempts to send a recording attendee into a Teams conference, the bot gets stranded in the virtual lobby or is expelled immediately by tenant DLP policies.
2. The $360 annual Copilot tax: To access native meeting summaries inside Microsoft Teams, organizations must purchase Microsoft 365 Copilot licenses priced at $30 per user per month ($360 annually per seat) with annual contractual commitments. For teams with dozens or hundreds of engineers and managers, this fee creates major budgetary friction.
3. Data residency and regulatory non-compliance: Financial institutions bound by FINRA/SEC regulations and healthcare organizations subject to HIPAA cannot permit consumer AI meeting bots to store raw conversational audio on third-party public cloud servers.
4. Proprietary cloud lock-in: Meeting transcripts generated inside Microsoft Teams live inside proprietary M365 databases, requiring painful copy-paste operations to integrate into an engineer’s personal Markdown knowledge base or Obsidian vault.

Desktop-native architecture: Operating beneath the tenant perimeter

Rather than attempting to negotiate API tokens with Microsoft 365 tenant controllers, Oyma operates entirely at the macOS operating system tier:

┌────────────────────────────────────────────────────────────────────────┐
│             MICROSOFT TEAMS WORKSTATION AUDIO ARCHITECTURE             │
├────────────────────────────────────────────────────────────────────────┤
│  M365 Cloud Tenant ───► Teams Client (macOS)                           │
│                              │                                         │
│                      CoreAudio Hardware Output                         │
│                              │                                         │
│                              ▼                                         │
│                    [ScreenCaptureKit Pipeline]                         │
│                              ▲                                         │
│                              │                                         │
│                       Local Microphone                                 │
│                              │                                         │
│                              ▼                                         │
│               [Apple Silicon Whisper Engine] (Local)                   │
│                              │                                         │
│                              ▼                                         │
│               Direct File Write: notes/meetings/*.md                   │
└────────────────────────────────────────────────────────────────────────┘

Because capture takes place inside Apple’s native ScreenCaptureKit framework:

• No artificial participant enters the Teams conference roster.
• No compliance alert flags an unvetted third-party service to IT administrators.
• Both the remote conference audio and your local spoken remarks are captured synchronously into local memory buffers with zero latency and zero cloud egress.

Proactive meeting detection on macOS

Oyma includes an automated CoreAudio listener in src/main/meeting/meeting-detector.ts that monitors macOS audio state transitions. Whenever the Microsoft Teams desktop executable or a browser-based Teams session activates a bidirectional audio route, Oyma identifies the VoIP session signature and displays an unobtrusive desktop banner:

Microsoft Teams conference active.
[Record Note]   [Ignore]

Recording remains strictly opt-in. The software never starts capturing without your affirmative click, guaranteeing total control over which enterprise discussions are recorded.

On-device Whisper transcription for enterprise confidentiality

When your Microsoft Teams conference finishes, Oyma feeds the captured audio directly into an optimized implementation of OpenAI Whisper running locally on your Mac’s Apple Silicon Neural Engine.

• Zero network egress: The entire speech-to-text pipeline runs offline on your local CPU/GPU/NPU silicon. No audio chunks travel across the corporate firewall or public internet.
• Multilingual fluency: Whisper handles 13 languages natively with automatic language identification, making it ideal for distributed international engineering divisions.
• Zero token metering: Transcribe unlimited hours of executive committees, backlog grooming sessions, and customer architecture syncs without per-minute transcription costs.

Comparison: Enterprise Teams capture approaches

Feature	Oyma on Mac	Microsoft 365 Copilot	Cloud AI Bots (Otter / Read)
Participant Bot in Call	None (Silent client capture)	None (Tenant service)	Yes (Visible external bot)
M365 Admin Consent Needed	No (Client workstation only)	Yes (Tenant approval)	Often blocked by tenant policy
Annual Per-Seat Price	Free during beta	$360 / user / year	$120–$240 / user / year
Audio Data Transmission	0% (100% on-device processing)	Microsoft Cloud data centers	Third-party cloud processing
File Format on Disk	Plain Markdown (.md)	Proprietary M365 cloud store	Vendor web dashboard
Vault Integration	Native Obsidian / file system	None (Manual copy-paste)	None (Manual export)

Enterprise Markdown record format

Every captured Teams call generates a clean, durable .md file written directly into your local vault directory:

---
title: "Core Infrastructure Migration: Quarterly Review"
date: 2026-10-08
platform: Microsoft Teams
attendees: ["David Vance", "Elena Rostova", "Marcus Chen"]
project: infra-migration
tags: [teams, enterprise, architecture, kubernetes]
---

## Executive Summary
Reviewed progress on the multi-region Kubernetes migration. Resolved database failover latency concerns and approved the target cutover timeline for European clusters.

## Architecture Decisions
- Database replication: Transitioning from asynchronous replication to semi-synchronous cross-region clustering for staging environments.
- Ingress gateway: Consolidating enterprise API routing through updated Envoy gateway clusters to unify TLS termination.
- Rollback SLA: Agreed on a strict 15-minute rollback window if synthetic traffic health dips below 99.95% during the production maintenance window.

## Action Items
- [ ] David: Finalize terraform manifest adjustments for the Dublin cluster by Friday
- [ ] Elena: Coordinate infosec vulnerability scanning on new container base images
- [ ] Marcus: Update the incident management runbook and circulate to on-call leads

Because these notes live as open files on your local drive, you retain complete ownership. You can search them with full-text search, navigate their relationships in graph view, and cross-reference them with your daily engineering log.

Auditing network security with the Egress Log

For corporate users who must prove to compliance officers or security operations teams that confidential discussions never leave the workstation, Oyma features a built-in Egress Log.

Every outbound HTTP request attempted by the desktop binary is logged and inspectable in real time. You can verify independently that during hours of Microsoft Teams recording and transcription, precisely zero bytes of audio or text leave your local Mac.

Capture your enterprise Microsoft Teams meetings with total security, zero bot friction, and local Markdown permanence.