Skip to main content
Browse a topic, then expand the question. If something does not work, see Troubleshooting. For a question missing from this page, open a GitHub issue.

​
Principles

No. Syntheo requires no account and has no subscription. All features are unlocked. The app is free software (MIT license).
Your own servers. Syntheo ships no model and uses no cloud AI service. In the settings, you enter the URL of a transcription server (STT) and the URL of a language model server (LLM), hosted on your computer or in your organization. See Transcription and AI.
For real-time transcription, Settings → Transcription recognizes two protocols:
  • “OpenAI · Realtime and batch”: any server that speaks the OpenAI Realtime protocol and /v1/audio/transcriptions, for example vLLM (with Voxtral Realtime), speaches, or faster-whisper-server.
  • “NVIDIA NIM · WebSocket”: an NVIDIA Speech NIM server deployed on your side.
For summaries and the chat, Settings → Synthesis accepts any OpenAI-compatible server exposing /v1/chat/completions and /v1/models, such as vLLM, SGLang, llama.cpp server, Ollama, LM Studio, TGI, or NIM LLM.
No. Syntheo captures your microphone and the system audio from your computer, without a meeting bot. See Record a meeting.
Yes, if your transcription and language model servers are reachable without internet: on the same computer (localhost) or on your local network. Syntheo needs no online service to work.

​
Data and privacy

No. The audio stays in memory long enough to be sent to your transcription server, then it is discarded. Syntheo creates no audio file and does not offer to retranscribe a meeting afterward. Only the transcript, your notes, and the summaries are saved.
Syntheo tries to reconnect. Since the audio is not kept, the part of the meeting spoken during the outage is missing from the transcript. A banner shows it with its times, for example “Transcription interrupted 14:32–14:35”. See Troubleshooting.
On your computer, in a local SQLite database. On Linux, the data folder is ~/.local/share/syntheo. Meeting content leaves your computer only to reach the STT and LLM servers you configured, and the webhooks you added. See Data and privacy.
No. Syntheo sends no telemetry, no crash reports, and no usage statistics to a publisher server.
The API key for your servers is optional. If you enter one, it is stored locally on this computer. It is never built into the app.
No. Syntheo has no cloud sync, no online sharing, and no team space. To pass a note on, export it with “Export” (PDF, TXT, Markdown, or Org) and send the file through the channel of your choice.

​
Meetings and notes

Yes. At the Synthesis step, show the transcript in the margin, then click “Edit” to correct the text and rename the speakers. See Notes.
Partly. Syntheo separates your voice (microphone) from the other participants’ (system audio). It also picks up the speakers your transcription server returns, if it provides them. There is no speaker analysis after the meeting.
Yes. Import a WAV, MP3, OGG, MP4, M4A, FLAC, WebM, or AAC file into an empty note. Syntheo sends it to your server without keeping a copy: through /v1/audio/transcriptions with the “OpenAI · Realtime and batch” engine, or into the /v1/realtime real-time stream with “NVIDIA NIM · WebSocket”. You can also import an SRT or VTT transcript. See Import a recording.
Choose your main language in Settings → Profile, “Language & Region” section. The languages actually transcribed depend on the model served by your transcription server: check its documentation. See Languages.

​
Platforms and updates

Linux (AppImage and .deb package) and Windows (signed .exe installer and .msi). No macOS version is published yet, and there is no mobile app. See Install the app.
Syntheo does not update itself. Download the latest version from the GitHub releases page and install it over the old one.
Yes, locally. The CLI and the local MCP server read the database on your computer. Webhooks push meeting events to your service while the app is open.