Nearly 3,000 listeners on an app that didn't carry our name

The international keynote speaker walks onto the stage of an Italian banking convention, in front of over three thousand people. The keynote is in English. Within a few minutes, almost the entire audience opens the event app, selects "Italian", and starts listening to the interpreter through their headphones.

The peak reaches 2,970 concurrent listeners on the Italian channel. Not on the Converso app, not on a second environment, not on a separate link: inside the technology partner's mobile app, with a white-label integration based on embeddable components.

For the end user, everything feels native to the event app. No second app to open, no extra step, no change of experience. For the partner, the integration stays simple: embedded UI, streaming, subtitles, reconnect, load handling, and real-time monitoring all managed by Converso.

The Challenge

The scenario was a stress test. The entire audience opens the translation simultaneously, inside a third-party app, live, at a high-visibility financial event. Zero direct control over the partner's native experience, over user devices, over the actual mobile network conditions in the room. Zero tolerance for delays, crashes, anomalies, or quality degradation. And zero Converso brand in front of the attendees: the banking group had commissioned the event to the technology partner, and the user experience had to remain entirely theirs.

Converso normally offers its own WebApp, already ready and optimised for multilingual events. In this case, however, the event already had its own official app developed by the technology partner. Adding a second touchpoint would have created friction: another app to open, another link to communicate, another possible point of confusion for the audience.

The better choice was to bring Converso quality inside the experience attendees were already using.

The Solution

Two iframes, zero streaming code in the partner's app

The entire user experience is built on two embeddable components that Converso provides as fully customisable iframes: one to listen to the interpreted audio in real time, one to read live subtitles synchronised with the audio. The partner integrated them pixel-perfect inside their app — colours, typography, the banking client's layout — without writing a single line of streaming code. WebRTC and WebSocket, mobile background persistence, reconnect logic, codec handling, network fallback: it all stays on Converso's side. The partner integrates the UI; we handle the streaming.

For whoever's building the app, this is the point: integrating means pasting two iframes. It doesn't mean becoming a real-time streaming engineer for an event.

The interpreter enters the platform from a browser

On the production side, the partner used Converso App Broadcaster, Converso's web-based broadcaster (all the ways to publish audio to the WebApp). No dedicated hardware in the production room, no client to install, no technical skills beyond what the partner already has. The interpreter's audio enters the Converso platform from a browser tab. From there, two parallel streams flow to each connected user: the interpreted audio onto the Converso WebApp embedded in the partner's app, and the live subtitles synchronised with that audio.

The critical moment: thousands of accesses in minutes

The most delicate moment was the start of the keynote. The audience opened the app almost simultaneously and the Italian channel climbed rapidly to nearly 3,000 concurrent listeners.

The ramp-up was concentrated in the first minutes: not traffic spread over the day, but thousands of activations while the speech was already underway. This is what made the case particularly interesting from an infrastructure standpoint.

Each attendee received interpreted audio and synchronised live subtitles. Overall, the infrastructure handled around 6,000 concurrent streams between audio and text data, keeping Converso's technical latency under 200 ms for the entire session. Those 200 ms measure the part that's on us — from the audio entering the Converso platform to its delivery to the attendee's smartphone — and at that level the technical delay isn't perceivable in the listening experience.

Latency under 200 ms is the minimum requirement of our daily work. What matters here is having delivered that same experience inside a partner's app, white-label, to an audience of nearly 3,000 concurrent listeners.

The ramp-up in the first minutes confirmed the infrastructure's resilience even in a concentrated-access scenario.

Control Room: present, invisible

In parallel, Converso's control room monitored latency, stream quality, media server health, live subtitle distribution, and client-side errors in real time — ready to intervene at the first sign of an anomaly. No intervention was needed. Neither the partner nor the audience ever saw this monitoring layer: which is exactly how it should be.

Why a human interpreter, here

There's a reason that at this event, in the booth, there was a person — not our RSAI engine. A banking convention of this stature puts the management on stage in front of their own audience: institutional register, financial jargon, sector irony, subtext, and the management's tone. It's exactly the kind of context where a human interpreter remains the right choice — because they understand the message, not just transfer it. They catch the nuance, know when an inflection matters more than a word, and recognise a sector joke before it becomes awkward in translation. For twenty-five years this has been at the heart of our craft, and here it was exactly the right call for an event of this nature.

Converso doesn't pick a side between AI and human. We pick the right interpreter for the right event — on the same infrastructure.

The Outcome

The peak was absorbed. The plateau was sustained. Audio and subtitles reached every connected attendee, cleanly, from the start to the end of the keynote, in a zero-tolerance financial context. Both the technology partner and the banking group expressed full satisfaction with stability and performance.

But the value of this deployment goes beyond a single event. The Converso engine is exactly the same whether the incoming voice is a human interpreter or our RSAI engine — Real-time Simultaneous AI Interpretation. For the partner that means one thing: the same app that today delivers traditional interpretation can, at any time, host AI interpretation. Without a line of additional code, without a single change to the integration. Same audio capture, same fanout, same listener experience.

Human interpretation today. RSAI tomorrow. Hybrid the day after. On the same infrastructure.

Already have an event app or a convention platform? Integrate live interpretation, subtitles, and RSAI white-label, without developing your own real-time infrastructure. → Integrate Converso into your app