v2.5.4

Some Microsoft Teams meetings deliver a single mixed audio stream, which left the bot with no way to tell speakers apart — those transcripts came back as "Speaker 1", "Speaker 2" and so on. Teams speaker detection now falls back through live captions and then on-screen observation, so a Teams meeting yields participant names one way or another.

⚡Improvements

• Teams transcripts carry participant names on meetings that deliver server-mixed audio, which previously produced generic speaker labels or an empty speaker timeline.
• Teams speech is matched to the speaker by audio timing rather than by arrival time, so consecutive speakers are no longer reported as talking over each other.
• Captions are read in the meeting's own language instead of being forced to English, which had suppressed them entirely on non-English tenants.

🐛Bug Fixes

• Teams recordings start clean instead of opening with several seconds of raw Teams interface before the bot tidies the view.
• A bot joining a Teams meeting that offers no chat continues immediately, instead of pausing twenty seconds first.
• A speaker turn that begins fractionally before the recording clock no longer produces a negative start time in the speaker timeline.

↩ All releases