An AI note taker converts meeting audio into a searchable transcript plus a structured summary (action items, decisions, key points). VexaScribe uses the upload model — no bot in your meeting, no participant alerts, works with any platform's recording (Zoom, Teams, Google Meet, Webex, phone calls). 30 min free, no signup.
TL;DR
An AI note taker converts meeting audio into a searchable transcript plus a structured summary (action items, decisions, key points). Two architectures: a bot that joins your call live, or an upload workflow that processes your recording after. VexaScribe uses the upload model — no bot in your meeting, no participant alerts, works with any platform's recording (Zoom, Teams, Google Meet, Webex, phone).
No bot, no calendar integration, no setup. Record your meeting the way you already do, upload after, get structured notes back in minutes.
MP3, M4A, MP4, MOV up to 5 GB. Or paste a URL (YouTube, Google Drive, Dropbox).
Whisper Large-v3 for transcription, pyannote-based diarization for who-said-what. 99 languages supported.
Full transcript plus a structured summary — action items, decisions, key points. Export to DOCX, TXT, SRT, VTT, JSON.
The platform your team uses shapes which AI note taker fits best. Here's the honest matrix.
When each wins: Native for teams already on Zoom Business+; Bot for real-time capture with teams already using Fireflies/Otter; Upload for ad-hoc meetings without pre-configured bot. Deep dive: Zoom transcription guide.
When each wins: Native if your org already pays for M365 Copilot; Bot for cross-tool workflows (CRM push); Upload for consultants or externals who aren't on your tenant. Deep dive: Teams transcription guide.
When each wins: Native for teams on Workspace Business Standard+; Extension for lightweight per-meeting use without a bot; Upload for cross-org meetings where you can't enable native. Deep dive: Google Meet transcription guide.
When each wins: Native for large enterprises already on Webex; Upload is the most reliable fit for Webex because third-party bot compatibility is inconsistent.
The single biggest architectural distinction. Most reviews skip it. Here's how they differ.
| Aspect | Bot-based | Upload-based |
|---|---|---|
| How it works | Bot joins your call as a participant | You record + upload after |
| Participant sees bot? | Yes (visible in participant list) | No |
| Works with any platform? | Only supported platforms | Any platform that produces a recording |
| Requires calendar/OAuth? | Usually | No |
| Privacy | Third party in your meeting | Your recording stays on your device until upload |
| Best for | Regular meeting workflows with team | Ad-hoc, sensitive, or multi-platform meetings |
| Examples | Fireflies, Otter, tl;dv, Fathom | VexaScribe, Sonix, HappyScribe |
Some teams actively avoid bot-based note takers. Reasons that come up in practice:
The upload-based model avoids all four: no bot in your meeting, no calendar OAuth, no third-party participant, no additional consent surface beyond your existing recording disclosure. You record with the tool you already use (Zoom's recorder, Teams, phone), then upload after.
For maximum control, self-hosted Whisper keeps the entire pipeline on your infrastructure — see the Whisper transcription guide for the DIY architecture.
VexaScribe positioning: no training on your data, recordings deleted after processing. If bot-free upload is what you need without running the pipeline yourself, we're the right shape.
Sales teams typically need more than a transcript: talk-ratio analysis, deal risk signals, and CRM sync (HubSpot, Salesforce, Pipedrive) so every call auto-attaches to the deal record.
Honest VexaScribe fit: we transcribe and summarize; we don't score deals or push to CRM natively. For full sales-workflow tooling, look at Modjo, Gong, or Chorus. Alternatively, use our transcript + a ChatGPT/Claude prompt for deal-risk analysis at a fraction of the cost.
Deep dive: Sales call transcription guide.
Therapists need HIPAA-adjacent handling and rigorous consent management. Session notes typically need to flow into an EHR.
Honest disclosure: VexaScribe is not HIPAA-certified. Use for consent-obtained recordings only, and never for regulated PHI where a BAA is required. For that, use a HIPAA-certified transcription service (Rev, dedicated medical scribes) or self-host Whisper.
Real workflow: session recording → transcript → therapist reviews notes → therapist uploads structured notes to EHR manually. The AI handles capture and structure; the clinician handles clinical judgment and compliance.
Speaker diarization is critical — you need to know exactly who said what for citation and analysis. Researchers also need to export cleanly into CAQDAS tools (NVivo, ATLAS.ti, MAXQDA).
Deep dive: Interview transcription guide.
Multi-client workflow: rename speakers per meeting, tag transcripts by client, export to DOCX for editable minutes you can send back to the client. The upload model is a natural fit — consultants often work across multiple client tenants where installing a bot isn't possible.
There's no universal best. Pick by architecture fit first (bot vs upload), platform second, price third.
| Tool | Architecture | Platforms | Free tier | Paid | Strength |
|---|---|---|---|---|---|
| Otter | Bot | Zoom, Meet, Teams | 300 min/mo (30-min file cap) | $16.99/mo Pro | Established, best live captions |
| Fireflies | Bot | Zoom, Meet, Teams, Webex | 800 min/mo (restricted) | $18/seat/mo | Deepest CRM integrations |
| VexaScribe | Upload | Any (recording-based) | 30 min, no signup, no card | From $9/mo | Bot-free, EU hosting, any platform |
| Fathom | Bot | Zoom, Meet, Teams | Free Pro (unlimited) | ~$19/seat/mo Team | Most generous free tier, sales-friendly |
| tl;dv | Bot | Zoom, Meet, Teams | Generous free tier | ~$20/seat/mo | Budget option |
| Krisp | On-device | Cross-platform desktop | Limited free tier | ~$8-$16/mo | Noise cancellation + notes |
| Notta | Upload + bot | Any + Zoom, Meet, Teams | 120 min/mo | ~$14.99/mo | Multi-language (58+) |
| Grain | Bot | Zoom, Meet, Teams | 20 recordings/mo | ~$19/seat/mo | Video clip highlights (sales) |
| Read.ai | Bot + extension | Zoom, Meet, Teams | Limited free tier | Pro tier | Meeting analytics overlay |
| Sembly | Bot | Zoom, Meet, Teams, Webex | 4 hrs/mo | ~$15/seat/mo | Task automation |
| Tactiq | Chrome extension | Meet, Zoom, Teams (browser) | 10 meetings/mo | ~$8/mo | No bot, extension-based |
Verdict: There's no universal best. Pick by architecture fit first (bot vs upload), platform second, price third. If you're on Zoom Business+ or M365 Copilot already, the native option is usually the cheapest incremental spend. If you want deep CRM integration, Fireflies. If you want bot-free with EU hosting, VexaScribe. If free-forever with a bot is fine, Fathom.
Under the hood, every AI note taker chains together four components:
Bot-based tools run this pipeline near-live (streaming ASR + progressive summarization). Upload-based tools run it in batch after you upload — same components, different scheduling.
Every vendor advertises a “free tier.” The honest breakdown of what you actually get:
| Tool | Free tier | Real-world limits |
|---|---|---|
| VexaScribe | 30 min free, no signup, no credit card | Full features on the free minutes |
| Otter | 300 min/month | 30-min per file cap, no export |
| Fireflies | 800 min/month | Feature restrictions, limited storage |
| tl;dv | Generous free tier (bot-based) | Advanced features gated |
| Whisper (open-source) | Totally free | Python + GPU required, DIY setup |
An AI note taker uses three technologies together: speech recognition (Whisper Large-v3 or similar) converts audio to text, speaker diarization (typically pyannote-based) separates who said what, and an LLM (GPT-4, Claude) generates a structured summary with action items and decisions. The whole pipeline runs in minutes for a 60-minute meeting.
There's no universal best — it depends on your workflow. If you want a bot in every meeting (regular team calls), Fireflies or Otter are strong. If you want no bot in your meetings (ad-hoc, sensitive, or privacy-focused), upload-based tools like VexaScribe work with any recording. Native options (Zoom AI Companion, Microsoft Copilot, Google Gemini) work best if your team is already on a paid plan for those platforms.
Yes, but the "free" varies. VexaScribe offers 30 min free with no signup or card. Otter gives 300 min/month free but caps each file at 30 min and blocks export. Fireflies gives 800 min/month free but with feature restrictions. Whisper open-source is totally free but requires Python and a GPU.
Yes. Upload-based AI note takers (VexaScribe, Sonix, HappyScribe) process your recording after the meeting instead of joining as a bot. No participant alert, no visible bot in the call, no third-party OAuth required. Trade-off: you don't get real-time capture — you upload after.
For Zoom: Fireflies (bot), Otter (bot), or Zoom's own AI Companion ($30/user/mo). For Microsoft Teams: Microsoft Copilot native (~$30/user/mo) or Fireflies. For Google Meet: Google's Gemini "Take notes for me" native (Workspace Business Standard+), Read.ai, or Tactiq Chrome extensions. For any platform via recording: VexaScribe (upload model).
Depends on the tool. VexaScribe doesn't train on user data and deletes recordings after processing. Most US-hosted tools (Fireflies, Otter) require additional data processing agreements for GDPR compliance. For sensitive/regulated data, self-hosted Whisper is the safest option.
Yes — Whisper (OpenAI) is open-source under MIT license. Combined with pyannote for speaker diarization and an open-source LLM (Llama, Mistral) for summarization, you can run a full AI note taker pipeline on your own infrastructure. Requires Python, GPU (or CPU with faster-whisper for slower processing), and engineering time to maintain.
Technical yes, compliance no. AI note takers transcribe therapy sessions accurately, but most consumer tools (including VexaScribe) are NOT HIPAA-certified. For clinical use, either use a HIPAA-certified transcription service (Rev, dedicated medical scribes) or self-host Whisper on your own infrastructure. Always obtain patient consent before recording.
How LLMs turn raw transcripts into structured meeting notes
Upload any recording (voice memo, lecture, phone call) — get structured notes with key points, decisions, and action items
Formal minutes with template structure (board, project, sales)
Every meeting type — bot vs upload, platform guides
Transcribe Zoom recordings — no bot needed
Transcribe Microsoft Teams recordings
Transcribe Google Meet recordings
How speaker diarization actually works
Self-hosted Whisper — the DIY AI note taker