Use faster-whisper 1.1.1 + wheeled av 13.1.0 (av 11 had no wheels and failed to compile on modern ffmpeg)

Dockerfile

@@ -13,13 +13,10 @@ ENV OMP_NUM_THREADS=1 \
     HUGGINGFACE_HUB_CACHE=/tmp/huggingface/hub
 
 # System libraries:
-#   ffmpeg / libsndfile1   - audio decode for whisper, soundfile, librosa
-#   git, build-essential   - building source-only wheels
-#   pkg-config + libav*-dev - required to compile PyAV (a faster-whisper dep)
+#   ffmpeg / libsndfile1 - audio decode for whisper, soundfile, librosa
+#   git, build-essential - building any source-only wheels
 RUN apt-get update && apt-get install -y --no-install-recommends \
-        ffmpeg git libsndfile1 build-essential pkg-config \
-        libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev \
-        libavfilter-dev libswscale-dev libswresample-dev \
+        ffmpeg git libsndfile1 build-essential \
     && rm -rf /var/lib/apt/lists/*
 
 # Install the CPU build of PyTorch first so the heavy CUDA wheel is never

requirements.txt

@@ -22,7 +22,8 @@ inputimeout==1.0.4
 evaluate==0.4.5
 accelerate==0.29.3
 huggingface_hub==0.20.3
-faster-whisper==0.10.0
+faster-whisper==1.1.1
+av==13.1.0
 edge-tts==6.1.2
 gunicorn
 python-dotenv