One person speaks — a sermon, a lesson, a meeting — and everyone hears and reads their own language on their own phone, over your WiFi, from a laptop in your building. No cloud platform in the middle, and no subscription to us — you bring your own translation provider key and see its per-minute cost on screen as you go.
People scan the code on the printed flyer or the projector and the page opens. Nothing to install — and the flyer carries a second code that joins your WiFi first, for guests who need it.
They pick their language from your service's list. Captions start immediately; a tap adds audio for earbuds.
The speaker just speaks. Every language streams live — phones, the projector caption wall, and room speakers, all at once.
A language counts as verified here only after a native speaker has listened to a real session and judged the translation. Everything else a provider offers should work, but check it with real speech before you rely on it.
| Gemini | OpenAI | Your own GPUopen stack | |
|---|---|---|---|
| English | ● | ○ | ○ |
| Spanish | ● | ○ | ○ |
| Portuguese | ● | ○ | ○ |
| French | ● | ○ | ○ |
| Haitian Creole | ◐weaker into English | –not offered | ○ |
| Russian | ● | ○ | ○ |
| Romanian | ●into English | –not offered | ○ |
| 70+ languages in all | 14 languages in all | 25 languages with voices |
Every pair of English, Spanish, Portuguese, French, Haitian Creole and Russian works in both directions. The one weak spot is a Haitian Creole speaker translated into English: usable, but less accurate than the rest. Romanian has so far been checked into English only, where it works well.
| From \ into | EN | ES | PT | FR | HT | RU | RO |
|---|---|---|---|---|---|---|---|
| English | · | ● | ● | ● | ● | ● | ○ |
| Spanish | ● | · | ● | ● | ● | ● | ○ |
| Portuguese | ● | ● | · | ● | ● | ● | ○ |
| French | ● | ● | ● | · | ● | ● | ○ |
| Haitian Creole | ◐ | ● | ● | ● | · | ● | ○ |
| Russian | ● | ● | ● | ● | ● | · | ○ |
| Romanian | ● | ○ | ○ | ○ | ○ | ○ | · |
The full lists, and how to choose a provider for each language, are in the languages guide.
Put every language up on a projector or TV, side by side, for people without a phone. Plug the screen into the laptop running Lithos and tick it on the This computer tab, or open the projector page in any smart TV or laptop browser on the same WiFi. From there you choose its languages, a column, row or grid layout, and when to show the join QR. It uses the same captions as the phones, so it adds nothing to your costs.
The Room audio graph follows microphones through rooms and languages to the speakers they feed. See a phone take the floor, spot an output that needs attention, and test the room before people arrive.
Lithos runs on your own hardware — a laptop at the sound desk, a mini-PC in a closet, a Raspberry Pi. Your congregation's words go from your microphone to their phones over your network. The only outside service involved is the AI translation provider you choose, with your own API key (how to get one) and their per-minute pricing in plain sight on the operator screen.
Free for churches, schools, and nonprofits. Lithos is source-available under the PolyForm Noncommercial license — congregations, charities, and classrooms use it at no charge, forever.
Using it commercially — conferences, paid events, service providers — requires a commercial license. Get in touch.
Desktop builds for macOS (signed and notarized), Windows, and Linux for both Intel and ARM — plus a headless server for a Raspberry Pi or Docker — are running in a small number of congregations before the public release. To be one of them, email support@lithos.community with your congregation and the languages you need; you get the installers by reply and a person to write to through your first Sunday. The getting started guide shows what running a service looks like. Every installer can be checked against our signing key.