← all SaaS
Self-host Otter.ai
Meeting transcription / AI notes ·
Category: AI / LLM tooling
Otter.ai records meetings, generates transcripts, and produces AI summaries with speaker attribution. The self-hostable replacement story revolves around OpenAI Whisper (now an open model) plus thin wrappers for diarization (who-spoke-when) and summarization (LLM call) — you assemble the pipeline yourself or pick a project that bundles it.
Otter.ai pricing anchor: Pro $16.99/user/mo (annual); Business $30/user/mo; Enterprise custom.
- GitHub
- ★ 49.5k · last commit 3d ago · 1196 open issues
- License
-
MIT - Setup time
- 20min — clone, build, test on a sample WAV
- Monthly cost
- $0 on a workstation; ~$50/mo for a cloud GPU box if running batch transcription regularly.
Migration sketch. `git clone https://github.com/ggerganov/whisper.cpp && make`. Download a model (`bash ./models/download-ggml-model.sh large-v3`). Transcribe: `./main -m models/ggml-large-v3.bin -f meeting.wav -ojf` produces JSON with timestamps. For speaker diarization, pair with `pyannote-audio` (Python, separate install) or use `whisper-diarization` which bundles both. Summarize the resulting transcript by piping into Ollama / a local LLM.
Good fit forEngineers who want a scriptable, shell-friendly transcription pipeline with no daemon to operate.
Weak atNo web UI, no live transcription out of the box; you build the meeting-recorder + speaker-attribution + summary glue.
- GitHub
- ★ 6.08k · last commit 11d ago · 355 open issues
- License
-
MIT - Setup time
- 5min installer (desktop app)
- Monthly cost
- $0 — runs on your laptop.
Migration sketch. Download the Vibe installer for macOS/Windows/Linux from thewh1teagle.github.io/vibe. Drop in an audio/video file or record system audio; Vibe runs Whisper locally and produces a transcript with timestamps + speaker hints. Summarization via the built-in LLM panel (point at Ollama or a hosted API). Otter recordings export as .mp3 or .mp4 — feed those into Vibe to backfill historical transcripts.
Good fit forSingle users who want a Whisper-powered desktop transcriber with a modern UI; no command line, no server.
Weak atSingle-user desktop only — no team workspaces or shared transcripts.
- GitHub
- ★ 19.1k · last commit 6d ago · 25 open issues
- License
-
MIT - Setup time
- 5min installer or `pip install buzz-captions`
- Monthly cost
- $0 desktop; ~$50/mo cloud GPU if you run a server.
Migration sketch. Buzz wraps Whisper.cpp with a desktop UI for transcription + live captioning. Install from chidiwilliams.com/buzz. Drop audio files in batch; outputs SRT/VTT/TXT/JSON. For server-mode live transcription, deploy `speaches-ai/speaches` (an OpenAI-API-compatible Whisper server) — your apps call `/v1/audio/transcriptions` exactly as they would call OpenAI.
Good fit forTeams that already rely on the OpenAI Audio API and want a drop-in self-hosted endpoint via Speaches plus a desktop tool for one-off jobs.
Weak atNo first-class meeting-recorder or speaker-diarization workflow bundled — pair with a separate diarization step.
In a terminal? npx os-alt otter-ai prints this table —
how the CLI works →