I'm experiencing an issue with Subtitles 1.8.2 (33) after upgrading my Mac mini to macOS 27.0 (build 26A428). Subtitles launches normally and system audio capture appears to work correctly, but no transcription is produced at all. As a result, no subtitle overlay appears. I tested the same version of Subtitles on my MacBook running macOS 26, where it works normally.
What I've tested on macOS 27.0:
Subtitles 1.8.2 (33)
System audio from Google Chrome
Dutch using the Multilingual model
English YouTube content using the English/Nemotron model
Single and multiple displays
Reset Overlay Position
Reset settings/model cache
Restarted macOS
Audio permissions checked
VAD / Skip non-speech tested
GUI and --headless mode
Audio capture itself appears to be working. In headless mode I get normal audio levels, VAD detects speech, and inference is running:
subtitles: 48000 Hz, 2 ch
listening. ctrl-C to stop.
Loading Nemotron · 560 ms…
Loading voice detection…
engine ready: Nemotron · 560 ms
[listening -21dB rtf 0.05 speech 45%]
[listening -12dB rtf 0.05 speech 45%]
[listening -14dB rtf 0.05 speech 45%]
[listening -13dB rtf 0.05 speech 45%]
However, no transcription text is ever printed to stdout.
I see the same behaviour with the Multilingual model for Dutch. Audio levels change normally and RTF is low (usually around 0.02–0.10), but no text is returned. For comparison, I installed/tested Subtitles on my MacBook running macOS 26, and transcription works normally there. So this does not appear to be an overlay positioning issue: even --headless receives audio and detects speech but produces no transcription on macOS 27.0.
This looks like a possible compatibility/regression issue with macOS 27.0 (26A428), potentially somewhere in the ASR/FluidAudio pipeline.
Happy to provide additional logs or run specific debug commands if useful.
I'm experiencing an issue with Subtitles 1.8.2 (33) after upgrading my Mac mini to macOS 27.0 (build 26A428). Subtitles launches normally and system audio capture appears to work correctly, but no transcription is produced at all. As a result, no subtitle overlay appears. I tested the same version of Subtitles on my MacBook running macOS 26, where it works normally.
What I've tested on macOS 27.0:
Subtitles 1.8.2 (33)
System audio from Google Chrome
Dutch using the Multilingual model
English YouTube content using the English/Nemotron model
Single and multiple displays
Reset Overlay Position
Reset settings/model cache
Restarted macOS
Audio permissions checked
VAD / Skip non-speech tested
GUI and --headless mode
Audio capture itself appears to be working. In headless mode I get normal audio levels, VAD detects speech, and inference is running:
subtitles: 48000 Hz, 2 ch
listening. ctrl-C to stop.
Loading Nemotron · 560 ms…
Loading voice detection…
engine ready: Nemotron · 560 ms
[listening -21dB rtf 0.05 speech 45%]
[listening -12dB rtf 0.05 speech 45%]
[listening -14dB rtf 0.05 speech 45%]
[listening -13dB rtf 0.05 speech 45%]
However, no transcription text is ever printed to stdout.
I see the same behaviour with the Multilingual model for Dutch. Audio levels change normally and RTF is low (usually around 0.02–0.10), but no text is returned. For comparison, I installed/tested Subtitles on my MacBook running macOS 26, and transcription works normally there. So this does not appear to be an overlay positioning issue: even --headless receives audio and detects speech but produces no transcription on macOS 27.0.
This looks like a possible compatibility/regression issue with macOS 27.0 (26A428), potentially somewhere in the ASR/FluidAudio pipeline.
Happy to provide additional logs or run specific debug commands if useful.