Live captions and translation at events, without venue internet.
A practical guide to running live captions and translation in an in-person event when you cannot rely on venue Wi-Fi.
ConferenceCaptioning is an on-device live captioning and translation system for in-person events. It runs on the organizer’s own Mac, iPad or iPhone, so captions and translation keep working without an internet connection. We recommend showing them on the room’s screens, so nobody needs a phone or a router. Attendees who want captions on their own phone can scan a QR code over a local network.
Why cloud tools struggle in event venues
Cloud captioning sends every word of audio to a data center and waits for the text to come back. That works until the network does not.
Shared Wi-Fi gets saturated
Hundreds of phones and laptops share one connection. Audio streaming to a remote server competes with all of them.
Delay compounds
Each word travels to the server and back. When the network is congested, captions fall seconds behind the speaker.
One failure stops everything
If the internet drops, cloud captions stop. A dedicated line reduces the risk, but you pay for it and still depend on it.
How offline event translation works
The AI runs on hardware you already own, using its GPU and Neural Engine or NPU, not a data center. We recommend putting captions on the room’s screens, so nobody has to use a phone.
- Bring the audio in. Feed your mixer output (Aux, USB, Dante or NDI) into a Mac, iPad or iPhone running ConferenceCaptioning. A Windows PC can use our Chrome extension.
- Transcribe and translate on the device. Speech recognition and translation happen locally, so no audio leaves the room.
- Show captions on the room’s screens. Send them to a display directly, or overlay them through ATEM, Roland or Tricaster as a downstream key, controlled from a Stream Deck or Bitfocus Companion. No router and no attendee devices are involved.
- Optional: add phones. If you want captions on personal devices too, bring a router, and attendees scan a QR code to read in their own language in a browser, with no app and no account.
- Keep a backup ready. A second Mac or iPad takes over automatically if the first one fails. See zero-downtime failover.
A real example: 10,000+ people in Bali
One speaker, talking in English. Four screens on stage, each showing captions in a different language at the same time.
An audience of more than 10,000 followed along on the screens in Mandarin, Bahasa Indonesia, Thai and English. Nobody needed a phone, a router or an app. The AI ran on the organizer’s own device.
A smaller example: at the NJEA Convention in August 2026, one computer acted as an offline AI server and overlaid live captions on the main screen through a video switcher.
Pre-event checklist
What to confirm before doors open, in the order that causes the fewest surprises.
- Decide where captions appear. Screens first. Confirm the connection from your device to the display or video switcher.
- Feed clean audio. Use a line feed from the mixer, not a room microphone, and check the level.
- Choose your languages. Decide which language goes on which screen. In Bali, each of four screens showed a different one.
- Keep the device powered and awake. Plug it in and turn off sleep before the event.
- Set up a backup device. A second Mac or iPad for automatic failover.
- Run a test with real speech. Play a recording of a speaker through the mixer and read the screens from the back of the room.
- Only if you want phones: bring a router, set a static IP address on the captioning device, and put the QR code on a slide or a table card.
Related setups: a local translation server on a Mac mini · captions in OBS · captions in ProPresenter · any slide deck
What “offline” means here, exactly
We are precise about this because event teams plan around it.
No internet needed for captions
- Speech recognition runs on the organizer’s device
- Translation runs on the organizer’s device
- Audio is not sent to a cloud service
- On screens, attendees need no network at all
A local network is needed only for phones
- Phones join a Wi-Fi network to reach the caption page
- That can be a router you bring, with no internet connection behind it
- The captioning device and the phones must be on the same network
When this is not the right fit
We would rather you pick the right tool.
Mostly remote viewers
If most people watch from other places, a cloud service reaches them without a device in the room.
You need a human interpreter or captioner
Legal, medical or regulated settings may require a certified person. AI captioning does not replace that.
You cannot place a device in the room
If there is no way to connect a device to your audio and screens, a browser-only service is simpler. See on-device vs cloud.
Measured, not claimed
We tested the on-device engines on a real NASA news conference at real-time speed. The results, the method and the data are public.
Best engine in that test: 97.2% word accuracy. Captions appear in under a second. Read the benchmark, and see how translation speed holds up in the multilingual translation benchmark.
Frequently asked questions
What is the best offline translation software for in-person conferences?
ConferenceCaptioning is built for this case. The speech recognition and translation run on the organizer’s own Mac, iPad or iPhone (or a Windows PC through a Chrome extension), so captions keep working when the venue internet is slow or down. We recommend showing them on the room’s screens, so nobody needs a phone. Whether it is the best fit depends on your room: if you need a human interpreter or remote viewers, a cloud service may suit you better.
How do you run live event subtitles without venue Wi-Fi?
Run the captioning on a device in the room, send the audio from your mixer into it, and show the captions on the room’s screens, either directly or through your video switcher. Nothing has to leave the building, so no internet line is needed for captions or translation. If you also want captions on attendees’ phones, add a router you bring and have them scan a QR code.
Do I need to bring a router?
Not if captions are on screens. We recommend showing them on the room’s screens so nobody has to use a phone. A router or access point is only needed if you also want attendees to read captions on their own phones.
Do attendees need internet to read the captions?
No. On screens they need nothing at all. On phones they need to join the local network the captioning device is on, which can be a router with no internet connection behind it. Captions open in the phone’s browser, with no app to install.
What happens if the captioning computer fails?
A second Mac or iPad can run as a backup. The display checks the primary every 0.5 seconds, and after the number of missed checks you choose (default 4, about 2 seconds) it switches to the backup without dropping attendee streams. You can set the switch to happen in as little as half a second, and an email alert is sent when failover happens.
How much does a dedicated event internet line cost?
Quotes vary by venue and event. Dedicated lines for large conferences commonly run from about $2,500 to $5,000 for the event. Because captioning and translation run on your own device, you do not need to buy extra bandwidth for them.
How accurate are offline captions?
In our published test on a live NASA news conference, run on a Mac with no internet needed, the best engine scored 97.2% word accuracy. The method, data and scorer are open, so you can check or repeat it.
Try it at your next event
Start free in about 15 seconds, or book a demo and we will walk through your room, audio setup and languages.