"The translation appears on my phone, so why not on the broadcast?"… COMEUP 2025 and the operations behind multilingual delivery

At an international conference, on-site attendees and online viewers may use language services on different screens. In its post-event account of COMEUP 2025, held at COEX in Seoul in December 2025, interpretation provider EventCAT said it supplied QR-accessible AI interpretation and captions across five stages, and live captions for the main-stage YouTube broadcast.

Access on a visitor’s phone and captions on a broadcast are separate outputs to check. Following the reference standard "Case Article Review", this column distinguishes the scope described by the supplier from testing suggestions for other events.

1. Separate output paths: check on-site access and broadcast captions independently

On-site attendees can use a QR code or URL to select a language on their own device. Online viewers need captions on the broadcast screen or a separate access route. Seeing a translation on a phone does not by itself establish that captions are working for online viewers too.

• Assign checks separately: At another event, the interpretation lead can check QR access and stage and language selection from an attendee’s seat, while the video lead checks caption display, placement and delay on the actual broadcast. The team can compare both results to identify the affected path if a problem occurs.
• Distinguish the case from a technical configuration: EventCAT’s current support documentation describes a way to composite a caption layer into a broadcast. Neither that document nor the supplier’s retrospective establishes the exact compositing method or wiring used at COMEUP 2025. The five-stage on-site service and the main-stage broadcast should also be described separately.

2. Recheck inputs and screens during session changes

When several stages run at once or speakers follow one another, teams need to check which speech appears on which screen. In a transition rehearsal for another event, stage staff can provide the interpretation lead with the next session’s language and speaker name, while the sound and video teams check that the inputs and outputs match the intended stage.

• Check previous displays and transition procedures: Look for a sentence left from the previous session or audio arriving from another stage, then clear a display or change an input when needed under the procedure agreed with the provider. Resetting the mixer or deleting the caption cache after every session is not a universal rule. Agreeing responsibilities and recovery methods in advance helps the team respond calmly to transition errors.

3. Hand over the speech and Q&A audio used for captions

For an interpretation service to process a presentation, speech from the correct stage must reach its input. EventCAT’s current on-site setup guidance describes receiving a clean signal from the sound mixer and checking its level. This is a configuration to consider for another event, not a record of the actual COMEUP wiring.

• Confirm the input scope with the sound team: The planner can arrange for the sound director and provider to review microphone types, available outputs, levels and network conditions together. If an audience question is not picked up by a microphone, the chair can repeat it into a microphone and the team can test that the repeated speech also reaches the interpretation input.
• Agree incident-response roles: Decide in advance who among the interpretation lead, video director and site lead checks and communicates a translation error or interruption. A fault may affect only one output, so record on-site and online status separately.

4. A four-stage protocol for multilingual delivery across media

The following suggestions are for events providing language services on personal devices in the venue and through an online broadcast. They are not a reconstruction of COMEUP 2025’s internal wiring or actual incident response.

Planner’s memo: check the same presentation again on each audience’s screen

When one presentation reaches both on-site and online audiences, their screens and language-selection options may differ. Planners should schedule time for the provider and video and sound teams to check their respective inputs and outputs.

Testing the same short passage on an actual attendee device and on the broadcast screen, and deciding who communicates what if an error occurs, helps teams run multilingual services more reliably.