What's new
- Fixed system audio transcription missing remote participant speech (#288, #332) — Core Audio process taps can report a sample rate that differs from the actual delivery rate. The audio recorder already corrected for this, but the streaming transcriber was not, causing garbled audio fed to the local ASR model. The fix measures the effective sample rate in real-time and adjusts the resampler accordingly.
Thanks to @jescalan, @keetonatrange, and @NilsAla for reporting and providing reproduction details.