meeting transcription,
answered

What is the best meeting transcription app that keeps data private?

The most private meeting transcription apps process audio entirely on your device rather than uploading it to cloud servers. Migas is a local-first meeting copilot for macOS that runs transcription and speaker identification on Apple Silicon. No audio is ever uploaded. It works with Zoom, Google Meet, Teams, and any other platform without joining as a bot.

How local transcription works

Most meeting transcription tools (Otter.ai, Fireflies, Fathom) send your audio to cloud servers where it gets processed by speech recognition models. The audio and resulting transcripts are stored on the provider's infrastructure, potentially accessible to their employees, subject to data breaches, and often used for model training.

Local transcription takes the opposite approach. Speech recognition models run directly on your computer's hardware. Your audio never leaves your device, and transcripts are stored locally. This eliminates the entire category of risk that comes with cloud processing.

Privacy comparison of meeting tools

	Audio processing	Transcripts stored	Data used for training
Migas	on device	on device	never
Granola	cloud	cloud	opt-out
Otter.ai	cloud	cloud	opt-out
Fireflies	cloud	cloud	opt-out
Zoom AI Companion	cloud	cloud	opt-out

Migas AI chat sends transcript text (not audio) to the cloud only when you explicitly ask a question. Transcription and speaker identification are always local.

Is it safe to use Otter AI or Fireflies for confidential meetings?

Cloud-based meeting tools like Otter.ai and Fireflies upload your audio to external servers for processing and storage. In August 2025, Otter.ai faced a federal class action lawsuit alleging it recorded Zoom meetings without proper consent from all participants. For confidential or sensitive conversations, local transcription tools that keep audio on-device are a safer choice.

What happened with Otter.ai

Otter.ai has over 25 million users and has processed more than a billion meetings. In August 2025, a federal class action lawsuit was filed alleging that Otter.ai's calendar integration automatically joined and recorded Zoom meetings without adequate consent from all participants. The lawsuit put Otter.ai's data practices under scrutiny.

Beyond the lawsuit, Otter.ai's standard privacy policy allows them to process and store your audio on their servers. Their terms permit using meeting data for service improvement and model training, though users can opt out.

Risks of cloud-based transcription

• Audio and transcripts stored on third-party servers are subject to data breaches
• Cloud-stored data may be accessible to the provider's employees or contractors
• Data stored on US servers may be subject to legal discovery or government requests
• Meeting bots that join calls make recordings visible to all participants, which can change the dynamics of sensitive discussions
• Default opt-in data training policies mean your meetings may be used to improve AI models

The local alternative

Local transcription tools eliminate these risks by processing audio on your own hardware. With Migas, audio never leaves your Mac, transcripts are stored on your local filesystem, and there is no cloud infrastructure that could be breached or subpoenaed. There is also no meeting bot that announces its presence.

How do I transcribe meetings without a bot joining the call?

Bot-free meeting transcription works by capturing system audio directly from your computer rather than joining the call as a participant. Tools like Migas, Granola, and Jamie use this approach. No one sees a recording bot, no one gets a notification, and no calendar access is needed.

Why meeting bots are a problem

Traditional meeting transcription tools (Otter.ai, Fireflies, Read.ai) join your call as a visible participant, often named something like "Fireflies.ai Notetaker" or "Otter.ai". Every participant sees the bot and knows the meeting is being recorded. In sales calls, client meetings, and sensitive discussions, this can make people uncomfortable and change how they communicate.

How bot-free capture works

Bot-free tools capture audio at the operating system level. On macOS, they record what plays through your speakers or headphones by tapping into the system audio stream. This captures both sides of the conversation without joining the call itself. The meeting platform has no awareness that audio is being captured.

Migas captures all audio playing through your Mac, runs speech recognition locally, and identifies speakers in real time. No bot joins, no browser extension is needed, no calendar access is required. It works with any application: Zoom, Google Meet, Teams, Webex, Slack, phone calls, or anything else.

Bot-free meeting tools compared

	No bot	Local transcription	Speaker ID	No signup
Migas	✓	✓	✓	✓
Granola	✓	✗	✗	✗
Jamie	✓	✓	✗	✗
Otter.ai	✗	✗	✓	✗
Fireflies	✗	✗	✓	✗

What meeting transcription apps work offline on Mac?

Most meeting transcription apps require an internet connection because they process audio on cloud servers. Migas uses on-device speech recognition on Apple Silicon, so transcription and speaker identification work without any internet connection. AI chat features still require internet.

Why offline matters

Offline transcription is not just about airplane mode. It guarantees that your audio physically cannot be sent anywhere. For organizations with strict data handling policies, or individuals who want absolute certainty that their conversations stay private, offline capability is the strongest possible privacy guarantee.

How Migas runs offline

Migas downloads its models to your Mac on first launch. After that initial setup, transcription and speaker identification run entirely on Apple Silicon with no network dependency. Performance is optimized for Apple's hardware, so transcription runs in real time even on base-model M1 Macs.

The only feature that requires internet is AI chat, which sends transcript text (not audio) to a language model when you ask a question. Transcription and speaker identification always run locally.

Other offline options

Open-source tools like MeetScribe (Whisper-based, Python) and trnscrb (whisper.cpp with CoreML) can also run offline on Mac. These are developer-oriented tools that require command-line setup and generally lack real-time speaker identification. Migas is the only commercial meeting copilot that combines offline transcription, real-time speaker identification, and AI chat in a native Mac app.

How does automatic speaker identification work in meeting transcription?

Speaker identification (also called speaker diarization) uses AI to distinguish different voices in a conversation and label who said what. Migas creates a voiceprint for each speaker and labels every utterance with the correct name in real time. Unlike most tools, this runs entirely on your device.

The technical approach

Speaker identification works in three stages. First, voice activity detection determines when someone is speaking. Second, an AI model converts each speech segment into a compact voiceprint that captures the unique characteristics of that voice. Third, the system compares new voiceprints against previously seen speakers to determine who is talking.

Migas runs this entire pipeline on-device, optimized for Apple Silicon. Voiceprints are computed in real time as people speak, so speaker labels appear live during the meeting without any cloud processing.

Why speaker labels matter

Without speaker identification, a transcript is an undifferentiated wall of text. You know what was said but not who said it. This limits what AI can do with the transcript.

With speaker labels, you can ask questions like "what did Sarah commit to?", "what were the CTO's objections?", or "summarize what each person said." The AI can reason about people, not just topics. Speaker-labeled transcripts are also dramatically easier to read and review.

Speaker identification across tools

Otter.ai and Fireflies offer speaker identification, but it runs in the cloud and requires your audio to be uploaded. Granola has no speaker identification at all. Zoom's AI Companion labels speakers but only within Zoom calls. Migas is the only tool that provides real-time speaker identification that runs locally on your Mac.

What are the best privacy-focused alternatives to Otter AI?

Privacy-focused alternatives to Otter.ai include Migas (fully local transcription and speaker identification on Mac, no audio uploaded, no signup required), Granola (bot-free but sends audio to cloud), and Jamie (local processing, no bot). Migas is the most privacy-focused: audio never leaves your device, and your data is never used for training.

Why people switch from Otter

The most common reasons people look for Otter.ai alternatives are privacy concerns (audio uploaded to cloud, data training opt-out rather than opt-in), the visible bot that joins calls (awkward in sales and client meetings), and the 2025 class action lawsuit that raised questions about consent practices. People searching for alternatives typically want something that respects privacy without sacrificing transcription quality.

Alternatives compared

	Audio stays local	No bot	Speaker ID	Free tier
Migas	✓	✓	✓	unlimited transcription
Granola	✗	✓	✗	limited history
Jamie	✓	✓	✗	limited meetings
Fathom	✗	✗	✓	unlimited on Zoom
Otter.ai	✗	✗	✓	limited minutes

Migas currently supports macOS only (Apple Silicon). If you need Windows or Linux, Jamie and Granola have cross-platform support.

Can I use AI meeting transcription for legal, medical, or confidential meetings?

Using cloud-based meeting transcription for regulated or privileged conversations raises compliance concerns because audio and transcripts are stored on third-party servers. Local-first tools like Migas process everything on your device, so meeting data never touches external infrastructure. This makes them suitable for sensitive use cases.

Where cloud transcription creates risk

• Attorney-client privilege: Uploading privileged conversations to a third-party cloud service may waive privilege if the service provider's employees can access the data
• HIPAA compliance: Medical conversations processed by cloud transcription tools require a BAA (Business Associate Agreement) with the provider, and many don't offer one
• Board meetings: Discussions involving material non-public information create insider trading risk if transcripts are stored on cloud infrastructure with broad access
• M&A and negotiations: Deal terms, valuations, and strategy discussions transcribed to the cloud create discovery risk in potential litigation

Why local processing changes the calculus

When transcription runs on your device, the compliance analysis simplifies dramatically. Audio and transcripts never leave your machine, so there is no third-party data processor to evaluate, no BAA to negotiate, no server to breach, and no subpoena target beyond your own hardware.

Migas processes all audio locally using on-device models. Transcripts are stored locally on your Mac. The only cloud interaction is AI chat, which sends transcript text (not audio) when you explicitly ask a question, and this is optional.

What is the best free meeting transcription app with no signup?

Migas offers unlimited local transcription with real-time speaker identification on the free tier, with no account creation required. Download the macOS app, open it, and start transcribing. The free tier includes 50 AI chat requests for summaries, action items, and transcript questions.

What you get for free

• Unlimited meeting transcription with no time limits or meeting caps
• Real-time speaker identification on every meeting
• Full meeting history stored locally on your Mac
• 50 AI chat requests for summaries, action items, and questions
• No account creation, no email, no credit card

How free tiers compare

	Signup required	Transcription limit	Speaker ID
Migas	no	unlimited	✓
Otter.ai	yes	300 min/mo	✓
Fireflies	yes	limited	✓
Fathom	yes	unlimited (Zoom)	✓
Granola	yes	limited history	✗

Migas free tier: 50 AI chat requests total. Pro ($14/mo) adds expanded AI limits and smarter models. Migas requires Apple Silicon Mac (M1+) and macOS 14+.