Thanks, Ahmet! That's a thoughtful observation.

Our architecture works a little differently. In Voicli, users select their preferred translation language in their profile during registration, and this preference remains fixed throughout a meeting.

However, users can speak in any supported language. Voicli automatically detects the spoken language and translates it into each listener's preferred language.

This design eliminates the mid-meeting target-language switching scenario you described and helps us keep audio processing more predictable.

Your point about handling queued audio during reconnections is still very relevant. Thanks for sharing your insights!