Transcribe with faster-whisper-base by default

Much faster than -small on a CPU-only host; TRANSCRIBE_MODEL still overrides it.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

docker-compose.yml

@@ -23,6 +23,7 @@ services:
       MS_CLIENT_SECRET: ${MS_CLIENT_SECRET:-}
       LYRICS_API_URL: http://127.0.0.1:9999                  # the bundled lyrics service (Lyrica, below, shares this container's network); remove this line and the service to skip it
       TRANSCRIBE_API_URL: http://127.0.0.1:8000              # the bundled speech-to-text service (whisper, below): makes a transcript for a podcast or audiobook on request. Remove this line and the service to skip it
+      TRANSCRIBE_MODEL: ${TRANSCRIBE_MODEL:-Systran/faster-whisper-base}   # smaller and about 3x faster than -small on a CPU-only host; set Systran/faster-whisper-small for better accuracy
       SIGNAL_API_URL: http://127.0.0.1:8085                  # the bundled Signal bridge (below, shares this container's network); remove this line and the service to skip Signal
       OPENAUDIBLE_URL: ${OPENAUDIBLE_URL:-}                # audiobooks straight from OpenAudible (e.g. http://172.16.16.106:8000); the username and password in Settings -> Audiobooks sign-in are OpenAudible's. Leave empty to go through the audiobook player instead
       PODCAST_URL: ${PODCAST_URL:-https://podcast.necloud.us/}