Meeting notes · checked 2026-10-19

Can I vibecode Otter.ai?

KINDA· weekend project

Otter is two products. One is a transcript library: audio in, speaker-labelled text out, with search, a summary and questions over everything. An agent builds that for one person in a weekend on faster-whisper, pyannote and SQLite. The other is the notetaker that joins Zoom, Teams and Meet from your calendar, plus the phone apps and shared workspaces. That part is meeting-bot infrastructure, and the open projects for it need servers and upkeep. The weekend version gives you uploaded or browser-recorded conversations, not the hands-off kind.

$16.99/moprice per seat

$204you'd save a year

weekendbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyAdvanced · Real infrastructure or OAuth work.

TimeA weekend; speaker labels and audio capture take the longest

Claude Code, Codex or CursorAny of them can run the prompt.
Python 3.12 and ffmpegInstall ffmpeg with Homebrew or your package manager.
A Hugging Face account and tokenSpeaker labels use pyannote models you must accept the licence for.
A machine that can run WhisperA recent Mac or a GPU is comfortable; a plain CPU works but is slow on long recordings.
A way to capture call audioChrome tab audio works for web calls; native apps on macOS need a virtual device such as BlackHole.

Before you buildCheck the free plan first: Otter Basic gives 300 minutes a month with live transcription, but caps each conversation at 30 minutes and keeps only the 25 most recent.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a self-hosted meeting transcriber to replace Otter.ai. Requirements:

- Python 3.12 + FastAPI with htmx pages. `uvicorn app:app --port 4860` serves http://localhost:4860, bound to
localhost; a docker-compose.yml runs the same thing. ffmpeg must be installed.
- Input, three ways: upload an audio or video file, record in the browser (getDisplayMedia for tab audio mixed with
the microphone, MediaRecorder, one chunk uploaded every 30 seconds), or drop a file into ./inbox/.
- Transcription: `faster-whisper` with word timestamps (WHISPER_MODEL, default medium; CPU int8, or CUDA if present).
A recording shows a rolling partial transcript every 30 seconds and gets a full pass when it stops.
- Speakers: `pyannote.audio` diarization merged with the words; renaming a speaker applies to the whole conversation.
It needs HF_TOKEN in .env and an accepted model licence; show that on the page when missing, never fail silently.
- Storage: SQLite at ./data/otter.db with conversations, segments (start, end, speaker, text) and an FTS5 index; audio
under ./data/audio/, deleted after KEEP_AUDIO_DAYS (default 30).
- Notes: a button sends the transcript to an OpenAI-compatible endpoint (LLM_BASE_URL, LLM_MODEL, LLM_API_KEY; Ollama
works) and stores a title, summary, decisions and action items with owners. A question box answers over one
conversation or all of them from the top FTS5 hits and cites timestamps.
- Playback page: click a sentence to jump the audio there, search across conversations, export TXT, SRT and DOCX
(`python-docx`).
- Single user. No accounts, no telemetry, local-first. Outbound calls: the model endpoint, and Hugging Face once for models.
- Out of scope: a bot that joins Zoom, Teams or Meet, calendar integration, mobile apps, sharing with other people.
- README: capturing a call's audio (Chrome tab audio, BlackHole on macOS), CPU against GPU speed, and a reminder to
tell participants you are recording.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Otter.ai.

Extend the transcriber in this folder. Requirements:

- Watch ./inbox/ and any extra folders listed in WATCH_DIRS with `watchdog`, and queue each new audio or video file once its size stops changing for 10 seconds.
- Process one file at a time in a background worker, show queue position and progress on the home page, and move finished files to ./inbox/done/.
- Retry a failed file once, then show the error and keep the file.
- Update the README.

Add speaker memory to the transcriber in this folder. Requirements:

- Store a voice embedding per named speaker from pyannote's embedding model, in SQLite.
- When a new conversation is diarized, suggest names for speakers whose embedding is close to a stored one and show the match score; accept or correct in one click.
- Never apply a suggestion without confirmation, and let me delete a stored voice.
- Update the README.

Add an action-item list to the transcriber in this folder. Requirements:

- Collect action items from every conversation into one page with owner, due date if stated, source conversation and a link to the exact timestamp.
- Let me tick items off and filter by owner; keep the state in SQLite.
- Provide a daily Markdown export to a path set in .env.
- Update the README.

What you lose

The notetaker that joins Zoom, Teams and Google Meet from your calendar by itself
The iOS and Android apps for recording on the go
Shared workspaces, shared speaker names and team permission controls
Salesforce, HubSpot and Zapier integrations, and auto-sync from Zoom or Dropbox
Live notes and captions inside the call

Why people still pay

The pull is capture without effort: the bot is in the call, the transcript is in your pocket, and nobody operates a server. Pro also lifts Basic's 30-minute cap per conversation and 300-minute pool, which is what makes regular use practical.

Moatintegrations execution polish

Do the math

Otter.ai or your own copy?

Seats

Over

Your usage

Count my time12h ×$/h

Otter.ai, 1 yr$204

Your own copy, 1 yr$44running + build session

You keep$160pays for itself in the first month

Estimate · Transcription runs on your machine at no cost. Summaries cost cents per meeting through an API, or nothing with Ollama; $6 covers a small VPS if you want it reachable from your phone. The build is token usage on your agent plan.

Leaving Otter.ai

Take your data, then cancel.

1 · Take your data out of Otter.ai

1Open the conversation, click the three dots at the top and choose Export.
2Pick a text format (TXT, DOCX, PDF or SRT), or turn on Audio to get an MP3; copying to the clipboard also works.
3Basic accounts can export text only as TXT; the other formats need a paid plan.
4Exporting many conversations at once is listed for Business and higher, so on Pro you export them one by one.

2 · Cancel Otter.ai

1Sign in as the account owner or an admin; only they can cancel.
2Go to Account Settings, then Billing, and choose Cancel plan.
3Give a reason, press Continue and confirm with Cancel subscription.
4Subscriptions bought through the Apple App Store or Google Play are cancelled in those stores instead.

The paid features run until the billing period ends; after that the workspace carries on as Basic with its limits.
Conversations stay in the account after you cancel or downgrade, but which ones you can open depends on the plan. The help center says an archived conversation can still be exported without upgrading.
Otter does not refund subscriptions under its Terms of Service, so cancel before the renewal date.
Export in DOCX, PDF or SRT while you are still on a paid plan, since Basic exports TXT only.
Whatever replaces the bot, tell participants when you record them.

Checked 2026-10-19 against Otter.ai's docs: 1, 2, 3, 4

The escape hatch

Don't feel like building it? These are already free.

ScriberrOpen sourceMIT self-hosted transcription server with speaker detection, and chat over transcripts through Ollama or an OpenAI-compatible API.★ 3.1k · active Sep 2026 · docker to self-host
WebVisit ↗MeetilyOpen sourceMIT desktop assistant that transcribes meetings locally and summarises with Ollama or your own API key; speaker labels are planned for the paid edition.★ 32k · active Sep 2026 · one-click install
macOS, WindowsVisit ↗BuzzOpen sourceMIT desktop app that transcribes and translates audio offline with Whisper; no bot, calendar or summaries.★ 22k · active Oct 2026 · one-click install
macOS, Windows, LinuxVisit ↗

All 5 free alternatives to Otter.ai ›

The numbers

Otter.ai pricing

Plan	Monthly	Annual, per month	What you get
Basic	$0	$0	300 monthly minutes (a pool shared with imports), 30 minutes per conversation, 3 lifetime file imports, 25 most recent conversations kept in history, 5 custom vocabulary terms, 1 concurrent meeting.
Pro	$16.99	$8.33	1,200 recording minutes and 10 file imports a month, up to 90 minutes per meeting, 2 concurrent meetings, unlimited history and storage, advanced AI workflows.
Business	$30	$19.99	Unlimited meetings and in-app recordings, 6,000 imported-file minutes a month, up to 4 hours per meeting, 3 concurrent meetings, activity logs, usage analytics, prioritised support.
Enterprise	custom	—	Everything in Business plus SSO and SCIM (100-user minimum), API and webhooks, custom integrations, HIPAA add-on.

Free tier: 300 minutes a month shared across recording and imports, 30 minutes per conversation, 3 imports ever, 25 most recent conversations.
Billing: Per user, monthly or annual (the page says annual saves 51% on Pro). Cards on the website plus in-app purchases through the Apple App Store or Google Play; PayPal is not accepted. Prices are in USD, and the Terms of Service say subscriptions are not refunded.
Hidden costs: Minutes are metered: Pro caps imports at 10 files and 1,200 imported minutes a month. Salesforce and HubSpot integrations are subject to user limits, live transcription over RTMP costs extra and HIPAA is a paid add-on.
Source: otter.ai

Also on the list

Granola$14/mo · Meeting notesYES

Read AI$19.75/mo · Meeting notesKINDA

Tactiq$12/mo · Meeting notesYES

Questions

Otter.ai, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

Basic is free: 300 transcription minutes a month, 30 minutes per conversation and 3 lifetime file imports.

Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: The notetaker that joins Zoom, Teams and Google Meet from your calendar by itself, The iOS and Android apps for recording on the go.

About $16.99 a month on the Pro plan per seat, $203.88 a year (checked 2026-10-19).

The notetaker that joins Zoom, Teams and Google Meet from your calendar by itself; The iOS and Android apps for recording on the go; Shared workspaces, shared speaker names and team permission controls; Salesforce, HubSpot and Zapier integrations, and auto-sync from Zoom or Dropbox; Live notes and captions inside the call. If any of those matter to you, keep paying.