Docker: FFmpeg with audio devices support (Pulse, Alsa) (#2586)

Signed-off-by: Viet Nguyen Duc <[email protected]>

Author: VietND96

.ffmpeg/Dockerfile

@@ -9,7 +9,7 @@ USER root
 # Install build tools
 #======================================
 ARG TOOLS_DEPS="autoconf automake cmake libfreetype6 gcc build-essential libtool make nasm pkg-config zlib1g-dev numactl \
-libnuma-dev libx11-dev libxcb-shm0 libxcb1-dev yasm git curl jq wget ca-certificates"
+libnuma-dev libx11-dev libxcb-shm0 libxcb1-dev libpulse-dev libasound2-dev yasm git curl jq wget ca-certificates"
 
 RUN apt-get update -qqy \
     && apt-get upgrade -yq \
@@ -66,6 +66,8 @@ RUN cd /usr/local/src \
     --enable-nonfree \
     --enable-libx264 \
     --enable-libxcb \
+    --enable-libpulse \
+    --enable-alsa \
     --enable-static \
     && make \
     && make install
@@ -80,7 +82,9 @@ COPY --from=builder /usr/local/bin/rclone /usr/local/bin/rclone
 
 RUN apt-get -qqy update \
     && apt-get -qqy --no-install-recommends install \
-    libx11-dev libxcb-shm0 libxcb1-dev \
+    libx11-dev libxcb-shm0 libxcb1-dev libpulse-dev libasound2-dev \
+    && apt-get -qqy update \
+    && apt-get -yq upgrade \
     && rm -rf /var/lib/apt/lists/* /var/cache/apt/*
 
 RUN ldd /usr/local/bin/ffmpeg \

ENV_VARIABLES.md

@@ -53,7 +53,7 @@
 | SE_HTTPS_PRIVATE_KEY | /opt/selenium/secrets/tls.key |  |  |
 | SE_ENABLE_TRACING | true |  |  |
 | SE_OTEL_EXPORTER_ENDPOINT |  |  |  |
-| SE_OTEL_SERVICE_NAME | selenium-router |  |  |
+| SE_OTEL_SERVICE_NAME | selenium-standalone-docker |  |  |
 | SE_OTEL_JVM_ARGS |  |  |  |
 | SE_OTEL_TRACES_EXPORTER | otlp |  |  |
 | SE_OTEL_JAVA_GLOBAL_AUTOCONFIGURE_ENABLED | true |  |  |
@@ -88,7 +88,7 @@
 | SE_REJECT_UNSUPPORTED_CAPS | false |  |  |
 | SE_NEW_SESSION_THREAD_POOL_SIZE |  |  |  |
 | SE_SESSION_REQUEST_TIMEOUT | 300 |  |  |
-| SE_SESSION_RETRY_INTERVAL | 5 |  |  |
+| SE_SESSION_RETRY_INTERVAL | 15 |  |  |
 | SE_HEALTHCHECK_INTERVAL | 120 |  |  |
 | SE_RELAX_CHECKS | true |  |  |
 | SE_SESSION_QUEUE_HOST |  |  |  |
@@ -119,3 +119,5 @@
 | SE_SUPERVISORD_LOG_FILE | /tmp/supervisord.log |  |  |
 | SE_SUPERVISORD_AUTO_RESTART | true |  |  |
 | SE_SUPERVISORD_START_RETRIES | 5 |  |  |
+| SE_RECORD_AUDIO | false | Flag to enable recording the audio source (default is Pulse Audio input) |  |
+| SE_AUDIO_SOURCE | -f pulse -ac 2 -i default | FFmpeg arguments to record the audio source |  |

Router/Dockerfile

@@ -13,7 +13,7 @@ USER ${SEL_UID}
 # In seconds, maps to "--session-request-timeout"
 ENV SE_SESSION_REQUEST_TIMEOUT="300" \
 # In seconds, maps to "--session-retry-interval"
-    SE_SESSION_RETRY_INTERVAL="5" \
+    SE_SESSION_RETRY_INTERVAL="15" \
     SE_OTEL_SERVICE_NAME="selenium-router"
 
 EXPOSE 4444

Video/Dockerfile

@@ -17,9 +17,11 @@ COPY --from=source /usr/local/bin/rclone /usr/local/bin/rclone
 
 RUN apt-get -qqy update \
     && apt-get -qqy --no-install-recommends install \
-    libx11-dev libxcb-shm0 libxcb1-dev \
+    libx11-dev libxcb-shm0 libxcb1-dev libpulse-dev libasound2-dev \
     x11-xserver-utils x11-utils \
     python3-pip \
+    && apt-get -qqy update \
+    && apt-get -yq upgrade \
     && pip install --upgrade --break-system-packages --no-cache-dir setuptools psutil \
     && rm -rf /var/lib/apt/lists/* /var/cache/apt/*
 
@@ -35,6 +37,8 @@ RUN ffmpeg -version \
 ENV DISPLAY_NUM="99" \
     DISPLAY_CONTAINER_NAME="selenium" \
     SE_RECORD_VIDEO="true" \
+    SE_RECORD_AUDIO="false" \
+    SE_AUDIO_SOURCE="-f pulse -ac 2 -i default" \
     SE_SERVER_PROTOCOL="http" \
     SE_VIDEO_POLL_INTERVAL="1" \
     SE_SCREEN_WIDTH="1920" \

Video/video.sh

@@ -204,14 +204,23 @@ function graceful_exit_force() {
   exit 0
 }
 
+if [ "${SE_RECORD_AUDIO,,}" = "true" ]; then
+  echo "$(date -u +"${ts_format}") [${process_name}] - Audio source arguments: ${SE_AUDIO_SOURCE}"
+else
+  SE_AUDIO_SOURCE=""
+fi
+
 if [[ "${VIDEO_UPLOAD_ENABLED}" != "true" ]] && [[ "${VIDEO_FILE_NAME}" != "auto" ]] && [[ -n "${VIDEO_FILE_NAME}" ]]; then
   trap graceful_exit SIGTERM SIGINT EXIT
   wait_for_display
   video_file="$VIDEO_FOLDER/$VIDEO_FILE_NAME"
   # exec replaces the video.sh process with ffmpeg, this makes easier to pass the process termination signal
   ffmpeg -hide_banner -loglevel warning -flags low_delay -threads 2 -fflags nobuffer+genpts -strict experimental -y -f x11grab \
-    -video_size ${VIDEO_SIZE} -r ${FRAME_RATE} -i ${DISPLAY} -codec:v ${CODEC} ${PRESET} -pix_fmt yuv420p "$video_file" &
-  wait $!
+    -video_size ${VIDEO_SIZE} -r ${FRAME_RATE} -i ${DISPLAY} ${SE_AUDIO_SOURCE} -codec:v ${CODEC} ${PRESET} -pix_fmt yuv420p "$video_file" &
+  FFMPEG_PID=$!
+  if ps -p $FFMPEG_PID >/dev/null; then
+    wait $FFMPEG_PID
+  fi
   wait_for_file_integrity
 
 else
@@ -243,8 +252,11 @@ else
       video_file="${VIDEO_FOLDER}/$video_file_name"
       echo "$(date -u +"${ts_format}") [${process_name}] - Starting to record video"
       ffmpeg -hide_banner -loglevel warning -flags low_delay -threads 2 -fflags nobuffer+genpts -strict experimental -y -f x11grab \
-        -video_size ${VIDEO_SIZE} -r ${FRAME_RATE} -i ${DISPLAY} -codec:v ${CODEC} ${PRESET} -pix_fmt yuv420p "$video_file" &
-      recording_started="true"
+        -video_size ${VIDEO_SIZE} -r ${FRAME_RATE} -i ${DISPLAY} ${SE_AUDIO_SOURCE} -codec:v ${CODEC} ${PRESET} -pix_fmt yuv420p "$video_file" &
+      FFMPEG_PID=$!
+      if ps -p $FFMPEG_PID >/dev/null; then
+        recording_started="true"
+      fi
       echo "$(date -u +"${ts_format}") [${process_name}] - Video recording started"
       sleep ${poll_interval}
     elif [[ "$session_id" != "$prev_session_id" && "$recording_started" = "true" ]]; then

scripts/generate_list_env_vars/description.yaml

@@ -360,3 +360,9 @@
 - name: SE_SUPERVISORD_START_RETRIES
   description: ''
   cli: ''
+- name: SE_RECORD_AUDIO
+  description: Flag to enable recording the audio source (default is Pulse Audio input)
+  cli: ''
+- name: SE_AUDIO_SOURCE
+  description: FFmpeg arguments to record the audio source
+  cli: ''

scripts/generate_list_env_vars/extract_env.py

@@ -6,6 +6,7 @@
 def extract_variables_from_shell_scripts(directory_path):
     variables = OrderedDict()
     for root, _, files in os.walk(directory_path):
+        files.sort()
         for file in files:
             if file.endswith(".sh"):
                 file_path = os.path.join(root, file)
@@ -23,6 +24,7 @@ def extract_variables_from_shell_scripts(directory_path):
 def extract_variables_from_dockerfiles(directory_path):
     variables = OrderedDict()
     for root, _, files in os.walk(directory_path):
+        files.sort()
         for file in files:
             if file.startswith("Dockerfile"):
                 file_path = os.path.join(root, file)
@@ -40,6 +42,7 @@ def extract_variables_from_dockerfiles(directory_path):
 def combine_dictionaries(dict1, dict2):
     combined_dict = OrderedDict(dict1)
     combined_dict.update(dict2)
+    combined_dict = dict(sorted(combined_dict.items()))
     return combined_dict
 
 def read_description_yaml(file_path):