Open-Source Meeting Transcription & Saving Solutions Research

This document evaluates open-source alternatives for real-time meeting transcription, with the capability to automatically save the transcription in a local Markdown (.md) format, using multiple audio API providers.

The evaluation is designed around the M2 Mac (16GB RAM) Apple Silicon hardware baseline and tailored to a non-IT professional / Business Analyst persona (prioritizing simple "click & run" GUI installers and high privacy/security compliance without command-line overhead).

📊 Comparison Matrix

Name (Link)	Stars	Latest Commit Date	Setup Overhead (Dễ/TB/Khó)	Pros & Cons	License & Cost	Key Core Features
Otter.ai (Baseline)	N/A	Proprietary	Dễ (Browser/App)	+ Highly polished UI and automated meeting bots. - Cloud-hosted (privacy risks); restrictive minute caps; high subscription pricing.	Proprietary / Free to Paid (starts at "$16.99 USD/month")	"300 monthly transcription minutes; 30 minutes per conversation" on Basic.
my-translator	`~1.2K`	2026-06-02	Dễ (Pre-compiled macOS dmg)	+ Supports multiple cloud STT/audio API providers; + Automatically saves aligned transcripts as `.md` locally per session; + Simple click-and-run GUI installer. - No automated calendar join bots.	MIT / Free (Pay-as-you-go API costs)	"Soniox · OpenAI Realtime · Qwen LiveTranslate Flash · Local MLX" engines.
meetily	`~12.5K`	2026-06-05	Dễ (Pre-compiled macOS dmg)	+ 100% local processing; extremely active community. - Lacks multi-provider audio transcription API support (transcription is local-only Whisper/Parakeet).	MIT / Free (PRO features are paid)	"Privacy-First AI Meeting Assistant... 100% local processing. no cloud required."
faster-whisper	`~23.6K`	2025-11-19	Khó (Terminal build, Python script)	+ High-speed, optimized local inference. - CLI/Library only; no GUI; no built-in `.md` saving or multi-provider API fallback.	MIT / Free	"up to 4–6x faster transcription compared to the original OpenAI Whisper implementation".
buzz	`~19.7K`	2026-06-15	Dễ (Pre-compiled macOS dmg)	+ Offline/GUI capability; supports external APIs. - No built-in dual-channel mixing (Zoom requires complex loopbacks).	MIT / Free	"transcribes and translates audio offline on your personal computer" with cloud API options.

🔍 Detailed Evaluation

1. my-translator (Recommended Winner)

Activity & Trust (70%): ~1.2K Stars. Extremely active commits (latest: June 2, 2026). Stable release cycle with built-in auto-updates.
Self-Hosting & Simplicity (30%): Dễ. Provides a "signed & notarized" macOS .dmg and Windows .exe out of the box. Zero terminal compiling needed.
API Providers & Transcription: Uniquely designed to support multiple real-time cloud and local APIs. It supports "Soniox · OpenAI Realtime · Qwen LiveTranslate Flash · Local MLX" directly. Users supply their own API keys, communicating directly with the endpoint with "no intermediary server involved" to protect privacy.
Saving Capabilities: Excellent. Transcripts are automatically "saved as .md files locally, per session", and a queue system ensures "source text is paired with its corresponding translation" to keep Markdown transcripts aligned.
Pros:
Multi-provider API support for transcription (OpenAI, Soniox, Qwen, Local MLX).
Auto-saves clean Markdown .md logs per session.
Captures Microphone, System Audio, or Both simultaneously (vital for Zoom/Teams calls).
Easy-to-use GUI desktop application with a minimal overlay.
Cons:
No calendar integrations or automated meeting join bots.
Local mode is only supported on Apple Silicon (though cloud APIs work on Intel/ARM alike).

2. meetily

Activity & Trust (70%): ~12.5K Stars. Highly popular, active commit velocity (latest: June 5, 2026).
Self-Hosting & Simplicity (30%): Dễ. Provides native .dmg and .exe installers.
API Providers & Transcription: Meetily supports multiple API providers for summarization (Claude, Groq, OpenRouter, OpenAI), but does not support multiple providers for the transcription itself. Transcription is performed "locally on your device using Whisper or Parakeet models". This is highly resource-intensive and drains battery on laptops.
Saving Capabilities: Saves meeting notes locally, but advanced exports and custom workflows are upsold to the "PRO Upgrade".
Pros:
Beautiful, modern UI with 100% local data sovereignty.
Local summaries are free (via Ollama).
Professional audio mixing.
Cons:
Cannot delegate live audio transcription to cloud endpoints like Deepgram or OpenAI API (strictly local models).
Advanced features are behind a paid "PRO" license.

3. faster-whisper

Activity & Trust (70%): ~23.6K Stars. Moderate activity (latest commit: Nov 19, 2025).
Self-Hosting & Simplicity (30%): Khó. No GUI, no standalone application. It is a developer-centric python library/CLI.
API Providers & Transcription: Only acts as a local transcription engine utilizing CTranslate2.
Saving Capabilities: None out of the box (requires writing custom scripts).
Pros:
High speed and low memory usage for local processing.
Cons:
Not user-friendly (requires developer setup).
Lacks meeting features, audio loopback capture, and native Markdown saving.

4. buzz

Activity & Trust (70%): ~19.7K Stars. Very active (latest commit: June 15, 2026).
Self-Hosting & Simplicity (30%): Dễ. Native installer provided.
API Providers & Transcription: Supports local Whisper and the OpenAI Whisper API.
Saving Capabilities: Exports to TXT, SRT, and VTT (does not format as a clean meeting Markdown document).
Pros:
Excellent offline desktop experience.
Supports using external OpenAI API keys.
Cons:
Lacks dual-channel loopback mixing: Cannot easily record both system audio (other participants) and microphone (yourself) simultaneously on macOS without complex third-party tools like BlackHole.

🏆 Recommended Winner: my-translator

For your objective of live meeting transcription + saving transcripts to .md files using multiple audio API providers, my-translator is the winner.

Why it fits:

True Multi-API Support: It lets you choose between Soniox, OpenAI Realtime, Alibaba Qwen LiveTranslate, and local Apple Silicon MLX.
Automatic Markdown Saving: Unlike other tools that export raw TXT or subtitle formats, my-translator is designed from the ground up to save aligned, structured bilingual or monolingual transcripts as .md files locally on your machine.
No-Setup Audio Mixing: Capturing both system audio and mic is notoriously difficult on macOS. my-translator includes native support for macOS ScreenCaptureKit and cpal to record "Both (System + Mic)" out-of-the-box without requiring complex virtual audio cable workarounds.
Click-and-Run: It provides a signed and notarized .dmg for Mac M2, aligning perfectly with a click-and-run workflow.