Meeting Transcription — Any Meeting, Transcript & Minutes

Depends on your workflow shape, honestly. If you want a bot in every calendar meeting with live notes and CRM push, dedicated note-takers lead: Otter ($8.33/mo annual), Fathom (generous free tier), Fireflies (deep CRM integrations). If you mostly have recorded meetings — Zoom/Teams/Meet exports, phone recordings of in-person meetings — an upload-first tool is cheaper and simpler: VexaScribe transcribes any recording at $2-$20/month with speaker labels on every plan, plus an opt-in Meeting Bot for the occasional live meeting (3× minute rate). If you're on a paid platform tier already, check the native option first: Teams and Meet both include transcription on eligible plans, with real limitations covered in our platform guides.

Several honest options. VexaScribe's first 30 minutes are free with no credit card — covers one typical meeting with full speaker labels and all export formats. Otter Basic gives 300 min/month recurring but caps each recording at 30 minutes. Fathom's free tier covers unlimited meetings if you accept a bot joining your calls. If your company already pays for Microsoft 365 or Google Workspace on the right tier, native Teams/Meet transcription is included — the catch is licensing gates and limited output formats (VTT or a Google Doc). For the full breakdown of every free option's real limit, see our free transcription comparison.

The workflow nobody's bot can help with — there's no video call to join. Record the meeting with a phone in the middle of the table (built-in voice recorder is fine; place it central, ask people to speak one at a time for the important parts), then upload the file. A 2-hour board meeting recorded as M4A is typically 60-120 MB — well within the 5 GB cap. Turn speaker diarization on: it will separate voices into Speaker 1/2/3 labels you rename once. One honest caveat: diarization accuracy drops when people talk over each other or sit far from the mic — a conference-mic or two phones at opposite ends of a long table meaningfully improves it.

Yes — that's the upload path. Record the meeting with the platform's own recorder (Zoom local/cloud recording, Teams recording, Meet recording) or a screen recorder, then upload the file afterward — or paste its Google Drive share link if that's where the recording landed. Nothing joins the call, no participant sees a bot, and the transcription costs regular 1× minutes instead of the 3× live-connection rate. This is the right path for client calls and interviews where a visible bot changes the dynamic, and the only path for recordings that already exist.

Two steps in one tool. After transcription, run the Meeting summary template: it produces a structured summary with decisions made, action items with owners (based on speaker attribution), and discussion topics — formatted as shareable minutes rather than a verbatim wall of text. You can also interrogate the transcript with AI Chat: ask "what did we decide about the Q3 budget?" and get an answer with timestamps validated against the source. For recurring formats (weekly standups, board meetings), export the DOCX and reuse your formatting.

On compressed meeting audio (Zoom/Teams/Meet recordings), Whisper Large-v3-class transcription reaches roughly 91-95% accuracy — noticeably below the 95-97% it achieves on clean studio audio, because meeting audio is compressed and multi-speaker. Accuracy drops further with crosstalk, strong accents, bad microphones, or conference-room echo. Practical planning: budget 5-10 minutes of proofreading per meeting hour to fix proper nouns and any misattributed speaker turns. Platform-native transcription (Teams, Meet, Zoom) generally runs below dedicated ASR models — usable for search, rough for sharing.

Yes, in two steps. Speaker diarization automatically detects voice changes and labels turns as Speaker 1, Speaker 2, and so on. You then rename each label once in the editor ("Maria — PM", "Client") and the names apply through the entire transcript and every export. Caveat that applies to every tool: when several people share one laptop microphone in a conference room, voices are genuinely hard to separate — per-participant audio (everyone joins from their own device) produces much better attribution. Platform-native transcription has an advantage here when everyone joins from named accounts: Teams tags turns with account names automatically.

Yes — this is exactly what upload-first transcription is for, and it's the case platform-native transcription can't handle at all (Teams and Meet transcription must be started during the meeting; it never applies retroactively). Find the recording file — Zoom saves locally to Documents/Zoom or to the cloud; Teams recordings live in the organizer's OneDrive/SharePoint; Meet recordings land in the organizer's Google Drive — and upload it, or paste the Drive share link directly. Any recording from any year works, up to 5 GB and 10 hours per file.

Recording laws vary by jurisdiction: some US states and many countries require all-party consent, others one-party. The safe, professional pattern regardless of jurisdiction: announce recording at the start of the meeting and get verbal agreement — which platform bots partially enforce by being visible, and which you should do manually for local recordings and in-person meetings. Company policy may add requirements beyond the law. This isn't legal advice — for regulated industries or cross-border calls, check with counsel.