How to Transcribe a Meeting Recording (Step-by-Step)

Transcribing a meeting recording sounds simple — drop the file into a tool, get the words back. The reality has more steps: there are recording consent laws to be aware of, platform-specific export options, multi-speaker diarization quirks, and a real choice between uploading after the fact versus running a real-time meeting bot. This guide walks the full pipeline honestly, including when our tool isn't the right fit.

First: the legal note (briefly, then move on)

Recording laws vary by jurisdiction. In the United States, federal law and most states are "one-party consent" — only one person on the call (which can be you) needs to know about the recording. Eleven states are "two-party" or "all-party" consent, including California, Florida, Illinois, Massachusetts, Pennsylvania, and Washington — every participant must agree.

Outside the US, laws vary widely. The EU generally requires all parties' consent under GDPR. The UK is similar. Many Asian and Latin American jurisdictions are stricter still.

The simple safe rule: announce at the start that you're recording. "Just a quick note — I'm recording this for my notes, is that okay with everyone?" This satisfies all-party consent laws and is good practice everywhere. Get a verbal yes on tape.

For specific legal questions (especially recording without disclosure, or recording where laws are ambiguous), consult an attorney. This is not legal advice; it's a starting orientation.

Step 1: get the recording

Zoom

1. Host: during the call, click Record at the bottom toolbar.
2. Choose Record on this Computer (saves locally as MP4 with separate M4A audio) or Record to the Cloud (Pro+ plans only; saves to your Zoom cloud account).
3. Stop recording at end of meeting; Zoom processes and saves the file.
4. Locate the file: usually in Documents/Zoom/ on Mac/Windows, organized by date.

If you're not the host, ask the host to record and share, or use a separate recording app on your machine (see below).

Google Meet

1. Recording requires Google Workspace Business Standard or higher.
2. During the call: More menu (three dots) → Record meeting → Start.
3. Files save to the host's Google Drive in a Meet Recordings folder.
4. Audio is bundled with video in MP4; extract audio with a free tool if you need MP3 for transcription.

Microsoft Teams

1. During the call: More actions (three dots) → Start recording.
2. Recording saves to OneDrive (1:1 calls) or SharePoint (channel meetings).
3. Teams now supports built-in transcription on Microsoft 365 plans — toggle Start transcription separately if you want both.

If you're not the host (or platform doesn't allow recording)

Use a separate recording app:

• macOS: QuickTime Player → File → New Audio Recording (records system audio plus mic via Loopback or BlackHole virtual audio device, free options exist)
• Windows: OBS Studio with a desktop audio capture (free, takes setup)
• Both: Krisp, Otter Pilot, or Fireflies bot if you want to add a transcription layer at the same time

Step 2: choose your transcription path

Two broad approaches:

Path A — upload the recorded file to a transcription tool

This is the right path for most users. Pros: works with any recording from any platform, no need to set up a bot, full control over which tool you use, no third-party bot in the meeting.

Steps:

1. Open a transcription tool. /convert/audio-to-markdown for Markdown output, TurboScribe for plain text + SRT, HappyScribe AI for highest accuracy.
2. Upload the audio file (MP3, M4A, WAV, MP4 video).
3. Wait for processing.
4. Review the transcript, fix obvious errors, edit speaker labels if needed.

For Markdown output specifically (which composes well with downstream AI tools), MDisBetter's audio-to-Markdown ships speaker labels, H2 section headers at topic shifts, and timestamps by default.

Path B — use a meeting bot for real-time transcription

The right path if you have recurring meetings and want hands-off automation. Otter.ai, Fireflies.ai, and a few others offer bots that auto-join your Zoom/Meet/Teams calls, transcribe in real time, and email the result.

Pros: zero post-meeting work, real-time captions visible during the call, integration with Salesforce/HubSpot/Slack/Notion. Best for sales teams and recurring multi-person meetings.

Cons: third-party bot is visibly in the meeting (some attendees object). Requires subscriptions for serious use. Most don't ship Markdown output. We don't ship a meeting bot — for this workflow, Otter or Fireflies is the right answer. We compare positioning in MDisBetter vs Otter.

Step 3: improve the transcript

Fix speaker labels

Most tools auto-label speakers as Speaker 1, Speaker 2, etc. Spend two minutes mapping these to actual names. Your future self (and any AI tool you give the transcript to) will thank you.

Spot-check the hardest moments

Diarization errors cluster around overlapping speech. Skim the transcript looking for sudden speaker changes mid-sentence — these are usually mislabels. Listen back to those moments and correct.

Add a meeting header

At the top of the transcript, add: meeting title, date, attendees, primary topic. This metadata is invaluable later when you're searching across many transcripts or feeding to an AI for summarization.

Step 4: extract action items with AI

This is where Markdown output really pays off. Paste the transcript into ChatGPT, Claude, or Gemini with a prompt like:

Below is a meeting transcript. Extract:
1. Decisions made (with the speaker who proposed them)
2. Action items (with owner, deadline if mentioned)
3. Open questions left unresolved
4. A 2-paragraph summary suitable for someone who wasn't there

Return each section in Markdown.

[paste transcript here]

Markdown-formatted transcripts (with H2 section breaks at topic shifts and labeled speakers) give the AI better navigation than flat text. The model can cite specific sections, attribute statements correctly, and pull action items with owners — all of which fail or degrade with plain-text input. We unpack the format-vs-quality argument in speech to text vs audio to Markdown.

Step 5: search and reuse

If you transcribe meetings regularly, build a searchable archive:

• Save each transcript with a consistent filename: YYYY-MM-DD_meeting-topic.md
• Drop them into Notion, Obsidian, or any Markdown-friendly note tool
• Use the tool's search to find quotes, action items, decisions across meetings

Markdown is the right archival format here too — readable in any text editor, queryable in any note app, immune to vendor lock-in if you change tools.

When a meeting bot beats file-upload

Meeting bots win when:

• You have recurring meetings on a calendar (the bot auto-joins everything)
• You want real-time captions visible during the call
• You need automatic CRM logging (Salesforce, HubSpot)
• You want shared team access to all meeting transcripts
• You don't want to remember to start recording or hit upload after

If three or more of these match, switch to Otter or Fireflies. The file-upload workflow gets tedious for high-frequency meeting users.

When file-upload beats a meeting bot

File-upload wins when:

• You record meetings sporadically rather than recurring
• You're on calls where a third-party bot would be unwelcome (sensitive client meetings, privacy-conscious participants)
• You want Markdown output for downstream AI processing
• You're already recording for compliance reasons (legal, medical) and just need transcription
• You don't want to pay for a meeting bot subscription
• You want to choose a different transcription tool per file (sometimes Whisper local, sometimes HappyScribe human, sometimes a quick web tool)

Multi-language meetings

If your meeting mixes English with another language (a common pattern in international teams), most tools handle it but with caveats:

• Whisper large-v3 handles code-switching the best — it can detect language shifts within a single segment
• HappyScribe supports 150+ languages but typically requires you to pick a primary language per file
• Otter and TurboScribe are English-strong; non-English support is more limited

For meetings primarily in one language with occasional excursions, Whisper local or HappyScribe AI work well. For truly bilingual meetings, expect some manual cleanup regardless of tool.

Recording quality tips for better transcripts

The audio quality you start with sets a hard ceiling on accuracy. Quick wins:

• Ask remote participants to use a headset rather than laptop speakers/mic
• For in-person meetings, use a centered conference-room mic rather than a laptop on one end of the table
• If a participant is in a noisy environment, ask them to mute when not speaking
• Record each Zoom participant's audio separately if possible (Zoom's local recording option supports this on some plans) — separate channels per speaker dramatically improves diarization

We unpack the audio-quality-vs-accuracy curve with real numbers in transcription accuracy by audio quality.

What about other formats in the same workflow?

Meeting prep often involves documents — agendas, prior decision logs, related research. Routing those through Markdown alongside the transcript creates a unified corpus you can feed to an AI for cross-document analysis ("what did we decide about X across the last six meetings, and how does it relate to the proposal in this PDF?"). See best free PDF to Markdown converters for the document side.

Privacy and retention

If you record meetings:

• Decide a retention policy (keep transcripts X months, then delete recordings/transcripts)
• Store transcripts somewhere with appropriate access controls — internal Notion, encrypted cloud, etc.
• For sensitive meetings (legal, medical, personnel), use Whisper local rather than uploading to cloud transcription
• If you use a meeting bot, check what the vendor does with the audio (training, retention) — paid plans of major vendors typically have stronger guarantees than free tiers

The honest summary

For one-off meeting recordings, upload to a transcription tool of your choice. For recurring meetings, especially with a team, pay for a meeting bot — Otter or Fireflies. For high-stakes or sensitive meetings, Whisper local plus careful retention policies. For Markdown-first AI workflows on uploaded files, MDisBetter gives you structured output by default. The mistake to avoid is building elaborate workflows around a tool that doesn't fit — try the file-upload path first, and only graduate to a meeting bot if frequency genuinely demands it.