Live AI Translation at Global Pastoralist Gathering and UNCCD COP17: How Omnio Is Bringing Language Access to Pastoralist and Indigenous Delegates in Mongolia

Omnio is a real-time AI translation platform for live events, and this month it's running at two of the biggest conferences on the international land and pastoralist policy calendar. From August 10 to 15, we're providing live conference translation at the Global Pastoralist Gathering in Ulaanbaatar, where around 300 delegates from 102 countries are meeting to talk about herding communities and the land they depend on. From August 17 to 28, we move into UNCCD COP17, where representatives from 197 Parties will spend two weeks negotiating the world's response to land degradation and drought, under the theme "Restoring Land. Restoring Hope."

Both events share a problem that rarely makes it into the official programme: most international conferences only budget interpretation for a handful of working languages. If a delegate's language isn't one of them, they get a seat, but not always a way to follow what's happening in it — and not always a way to be understood when they speak. That gap falls hardest on exactly the people whose input these conferences most need: pastoralist communities, indigenous delegates, and smaller national delegations who don't arrive with their own interpreters.

That's the specific problem Omnio is built to solve.

The platform combines automatic speech recognition, neural machine translation, and text-to-speech into one pipeline. A speaker talks into a podium or handheld mic; within about 4.8 seconds, the audience hears it back — as live translated audio and on-screen captions — in whichever of 100+ languages they need, on their own phone. There's no interpreter booth to book, no headset to hand out, no app to download. Attendees open a link, pick a language, and they're in the session. Accuracy holds at 95 to 97 percent even across the kind of long, multi-day summits where human interpreters start to fatigue by the third or fourth day.

For organizers, this also means the traditional cost structure of interpretation — where language services alone can run 20 to 40 percent of an event's budget — doesn't have to be the ceiling on how many languages you can support. For attendees, it means the choice isn't "wait for an interpreter to be available" versus "sit it out." Someone speaking a language no conference has ever had a budget line for can still follow the plenary, still use the platform's in-app 'Ask a Question' feature, and still be part of the record.

We didn't build Omnio to replace professional interpreters, and at events with the budget for full interpreter teams, it doesn't need to. We built it for the much more common situation: an event with dozens of language communities in the room and interpretation for maybe six of them. That's most conferences. It's certainly most conferences where pastoralist and indigenous delegates are in the room.

What we're watching for over these two weeks isn't really a technology story. It's whether a herder from a region with no official conference language, or a youth delegate from a country that rarely gets floor time, ends up in the transcript — asking a question, making a point, changing how a session goes. That's what multilingual accessibility is actually for. Not a nicer captions bar. A wider set of people whose knowledge counts.

We'll be posting from the ground through both events — what languages we're translating, who we're hearing from, and how live AI interpretation is showing up in the actual negotiations at COP17. If you're organizing a conference where language access shouldn't be limited by budget, you may read more about omnio at https://acaiberry.io/products/omnio