Select the speech microphones
Send the lectern, headset, host and panel microphones that carry the spoken program. Exclude instruments, playback, effects returns and room or audience microphones from this bus.
A microphone that carries both speech and singing needs operator attention during songs. We make no accuracy promise for speech over a music bed. Routing a clean feed is the starting point; it is not a measured accuracy guarantee.
Create a dedicated caption bus
Use a post-fader aux, or a matrix fed from a speech group if the console supports that routing. A post-fader send follows the speech channel’s fader and mute.
Avoid a pre-fader send for this purpose: a microphone that is muted in the house mix could remain audible on another bus. Verify the actual console routing rather than relying on the bus name. Save it in the console scene for the next event.
Connect the bus to the computer
If the console exposes USB audio, route the speech bus to a USB output channel. Otherwise use a dedicated console output into an audio interface’s line input. Select the intended input device and channel on the caption computer.
Keep the route simple enough to trace: microphone → speech bus → computer input. The room loudspeakers and full stream mix should not be the source. Check speech levels without clipping and listen to the feed itself.
Keep speech loud enough
Set the gain so normal speech sits around the console’s nominal level, without clipping. Speech that is too quiet is harder to recognize. Keep the microphone close and steady, and listen to the feed itself.
Watch for speakers who lower their voice or step away from the microphone. Check each person at their normal speaking volume before the event and monitor the level during it.
One voice at a time
Overlapping speakers reduce recognition accuracy. Ask speakers to take turns. Interpreters should wait for a pause before speaking; on panels, keep one microphone open at a time where possible.
These are setup recommendations, not measured accuracy claims or a guarantee of results. Keep music out of the speech bus; we make no accuracy promise for speech over a music bed.
Before anyone walks on stage
Check each speech microphone individually. Confirm that its fader and mute affect the caption feed. Play music through the house system and confirm it is absent from the speech bus.
During the pilot, test French, English and mixed speech on that route and verify the caption mute control. Screens and control names are still subject to pilot validation. Read the hardware requirements as well: clean audio and a suitable computer are both required.