make it work under macos

e2e_video_modes.py

@@ -15,8 +15,13 @@ SOURCE_FILE : str = os.path.join(os.path.dirname(os.path.abspath(__file__)), '.a
 def is_windows() -> bool:
 	return platform.system().lower() == 'windows'
 
+def is_macos() -> bool:
+	return platform.system().lower() == 'darwin'
+
 if is_windows():
 	VIDEO_FILE : str = 'C:\\Users\\info\\Downloads\\face8k.mp4'
+elif is_macos():
+	VIDEO_FILE : str = '/Users/henry/Downloads/copy_face_instant.mp4'
 else:
 	VIDEO_FILE : str = '/home/henry/Documents/examples/download.mp4'
 
@@ -27,7 +32,7 @@ def safe_print(text : str) -> None:
 	except UnicodeEncodeError:
 		print(text.encode('ascii', errors='replace').decode('ascii'))
 
-MODES =\
+_ALL_MODES =\
 [
 	'whip-mediamtx',
 	'whip-python',
@@ -38,12 +43,14 @@ MODES =\
 	'ws-mjpeg'
 ]
 
+MODES = [ m for m in _ALL_MODES if not (is_macos() and m == 'whip-mediamtx') ]
+
 
 def start_api() -> subprocess.Popen:
 	env = os.environ.copy()
 	python_cmd = 'python' if is_windows() else 'python3'
 
-	if not is_windows():
+	if not is_windows() and not is_macos():
 		env['LD_LIBRARY_PATH'] = '/home/henry/local/lib:' + env.get('LD_LIBRARY_PATH', '')
 
 	proc = subprocess.Popen(
@@ -103,12 +110,32 @@ def kill_port_windows(port : int) -> None:
 				subprocess.run([ 'taskkill', '/F', '/PID', pid ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
 
 
+def kill_port_macos(port : int) -> None:
+	pids = set()
+
+	for proto in [ 'tcp', 'udp' ]:
+		result = subprocess.run(
+			[ 'lsof', '-ti', proto + ':' + str(port) ],
+			capture_output = True, text = True
+		)
+
+		for pid in result.stdout.split():
+			if pid.isdigit():
+				pids.add(pid)
+
+	for pid in pids:
+		subprocess.run([ 'kill', '-9', pid ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
+
+
 def kill_stale() -> None:
 	ports = [ API_PORT, 8889, 8189, 9997, 8890, 8891, 8892 ]
 
 	if is_windows():
 		for port in ports:
 			kill_port_windows(port)
+	elif is_macos():
+		for port in ports:
+			kill_port_macos(port)
 	else:
 		subprocess.run([ 'fuser', '-k', str(API_PORT) + '/tcp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
 		subprocess.run([ 'fuser', '-k', '8889/tcp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)

facefusion/apis/core.py

@@ -13,14 +13,14 @@ from facefusion.apis.endpoints.metrics import get_metrics, websocket_metrics
 from facefusion.apis.endpoints.ping import websocket_ping
 from facefusion.apis.endpoints.session import create_session, destroy_session, get_session, refresh_session
 from facefusion.apis.endpoints.state import get_state, set_state
-from facefusion.common_helper import is_windows
+from facefusion.common_helper import is_linux, is_windows
 from facefusion.apis.endpoints.stream import websocket_stream, websocket_stream_audio, websocket_stream_live, websocket_stream_mjpeg, websocket_stream_rtc, websocket_stream_rtc_relay, websocket_stream_whip, websocket_stream_whip_dc, websocket_stream_whip_py
 from facefusion.apis.middlewares.session import create_session_guard
 
 
 @asynccontextmanager
 async def lifespan(app : Starlette) -> AsyncGenerator[None, None]:
-	if not is_windows():
+	if is_linux():
 		mediamtx.start()
 		mediamtx.wait_for_ready()
 
@@ -45,7 +45,7 @@ async def lifespan(app : Starlette) -> AsyncGenerator[None, None]:
 
 	yield
 
-	if not is_windows():
+	if is_linux():
 		mediamtx.stop()
 
 	try:

facefusion/apis/endpoints/stream.py

@@ -12,6 +12,7 @@ from starlette.websockets import WebSocket
 
 from facefusion import logger, session_context, session_manager, state_manager
 from facefusion.common_helper import is_windows
+from facefusion.apis.stream_helper import STREAM_AUDIO_RATE
 from facefusion.apis.api_helper import get_sec_websocket_protocol
 from facefusion.apis.session_helper import extract_access_token
 from facefusion import mediamtx
@@ -158,7 +159,24 @@ async def websocket_stream_whip(websocket : WebSocket) -> None:
 	await websocket.close()
 
 
-def run_fmp4_pipeline(latest_frame_holder : list, lock : threading.Lock, stop_event : threading.Event, output_chunks : List[bytes], output_lock : threading.Lock, audio_write_fd_holder : list) -> None:
+def run_audio_silence_feeder(audio_write_fd_holder : list, stop_event : threading.Event, audio_active_event : threading.Event) -> None:
+	frame_bytes = STREAM_AUDIO_RATE // 50 * 2 * 2
+	silence = b'\x00' * frame_bytes
+
+	while not stop_event.is_set():
+		if not audio_active_event.is_set():
+			fd = audio_write_fd_holder[0]
+
+			if fd > 0:
+				try:
+					_os.write(fd, silence)
+				except OSError:
+					break
+
+		time.sleep(0.02)
+
+
+def run_fmp4_pipeline(latest_frame_holder : list, lock : threading.Lock, stop_event : threading.Event, output_chunks : List[bytes], output_lock : threading.Lock, audio_write_fd_holder : list, audio_active_event : threading.Event) -> None:
 	encoder = None
 	audio_write_fd = -1
 	reader_thread = None
@@ -194,6 +212,8 @@ def run_fmp4_pipeline(latest_frame_holder : list, lock : threading.Lock, stop_ev
 					audio_write_fd_holder[0] = audio_write_fd
 					reader_thread = threading.Thread(target = read_fmp4_output, args = (encoder, output_chunks, output_lock), daemon = True)
 					reader_thread.start()
+					silence_thread = threading.Thread(target = run_audio_silence_feeder, args = (audio_write_fd_holder, stop_event, audio_active_event), daemon = True)
+					silence_thread.start()
 					logger.info('fmp4 encoder started ' + str(width) + 'x' + str(height), __name__)
 
 				feed_whip_frame(encoder, temp_vision_frame)
@@ -222,7 +242,8 @@ async def websocket_stream_live(websocket : WebSocket) -> None:
 		lock = threading.Lock()
 		output_lock = threading.Lock()
 		stop_event = threading.Event()
-		worker = threading.Thread(target = run_fmp4_pipeline, args = (latest_frame_holder, lock, stop_event, output_chunks, output_lock, audio_write_fd_holder), daemon = True)
+		audio_active_event = threading.Event()
+		worker = threading.Thread(target = run_fmp4_pipeline, args = (latest_frame_holder, lock, stop_event, output_chunks, output_lock, audio_write_fd_holder, audio_active_event), daemon = True)
 		worker.start()
 
 		try:
@@ -245,6 +266,7 @@ async def websocket_stream_live(websocket : WebSocket) -> None:
 								latest_frame_holder[0] = frame
 
 					if data[:2] != JPEG_MAGIC and audio_write_fd_holder[0] > 0:
+						audio_active_event.set()
 						feed_whip_audio(audio_write_fd_holder[0], data)
 
 		except Exception as exception:

facefusion/mediamtx.py

@@ -6,6 +6,8 @@ from typing import Optional
 
 import httpx
 
+from facefusion.common_helper import is_linux
+
 
 MEDIAMTX_WHIP_PORT : int = 8889
 MEDIAMTX_API_PORT : int = 9997
@@ -56,9 +58,10 @@ def stop() -> None:
 
 
 def stop_stale() -> None:
-	subprocess.run([ 'fuser', '-k', str(MEDIAMTX_WHIP_PORT) + '/tcp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
-	subprocess.run([ 'fuser', '-k', '8189/udp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
-	subprocess.run([ 'fuser', '-k', str(MEDIAMTX_API_PORT) + '/tcp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
+	if is_linux():
+		subprocess.run([ 'fuser', '-k', str(MEDIAMTX_WHIP_PORT) + '/tcp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
+		subprocess.run([ 'fuser', '-k', '8189/udp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
+		subprocess.run([ 'fuser', '-k', str(MEDIAMTX_API_PORT) + '/tcp' ], stdout = subprocess.DEVNULL, stderr = subprocess.DEVNULL)
 	time.sleep(1)
 
 

facefusion/rtc.py

@@ -7,7 +7,7 @@ from http.server import BaseHTTPRequestHandler, HTTPServer
 from typing import Dict, List, Optional, TypeAlias
 
 from facefusion import logger
-from facefusion.common_helper import is_windows
+from facefusion.common_helper import is_macos, is_windows
 
 RtcLib : TypeAlias = ctypes.CDLL
 WHEP_PORT : int = 8892
@@ -87,6 +87,8 @@ def find_library() -> Optional[str]:
 
 	if is_windows():
 		return os.path.join(bin_dir, 'windows-x64-openssl-h264-vp8-av1-opus-datachannel-0.24.1.dll')
+	if is_macos():
+		return os.path.join(bin_dir, 'macos-universal-openssl-h264-vp8-av1-opus-libdatachannel-0.24.1.dylib')
 	return os.path.join(bin_dir, 'linux-x64-openssl-h264-vp8-av1-opus-libdatachannel-0.24.1.so')
 
 
@@ -206,6 +208,8 @@ next_rtp_port : int = 16000
 
 
 def create_session(stream_path : str) -> None:
+	global video_frame_count
+	video_frame_count = 0
 	sessions[stream_path] = {'viewers': [], 'tracks': [], 'rtp_port': 0, 'rtp_fd': None}
 
 
@@ -288,6 +292,7 @@ def send_audio_to_viewers(stream_path : str, opus_data : bytes) -> None:
 
 
 send_start_time : float = 0
+video_frame_count : int = 0
 audio_pts : int = 0
 opus_enc = None
 audio_buffer : bytearray = bytearray()
@@ -296,7 +301,7 @@ OPUS_FRAME_SAMPLES : int = 960
 
 
 def send_to_viewers(stream_path : str, data : bytes) -> None:
-	global send_start_time
+	global video_frame_count
 
 	session = sessions.get(stream_path)
 
@@ -308,11 +313,8 @@ def send_to_viewers(stream_path : str, data : bytes) -> None:
 	if not viewers:
 		return
 
-	if send_start_time == 0:
-		send_start_time = _time.monotonic()
-
-	elapsed = _time.monotonic() - send_start_time
-	timestamp = int(elapsed * 90000) & 0xFFFFFFFF
+	timestamp = int(video_frame_count * 3000) & 0xFFFFFFFF
+	video_frame_count += 1
 	buf = ctypes.create_string_buffer(data)
 	data_len = len(data)
 

test_stream.html

@@ -260,7 +260,7 @@
     </div>
 
     <div class="video-area">
-      <video id="outputVideo" autoplay playsinline></video>
+      <video id="outputVideo" autoplay muted playsinline></video>
     </div>
     <video id="inputVideo" autoplay muted playsinline style="display:none;"></video>
 
@@ -415,7 +415,7 @@ function startStats() {
           prevBytes = bytes;
           prevFrames = frames;
           prevStatsTime = now;
-          setStat('statFrames', report.framesReceived || 0);
+          setStat('statFrames', report.framesDecoded || 0);
 
           if (report.frameWidth && report.frameHeight) {
             setStat('statResolution', report.frameWidth + 'x' + report.frameHeight);
@@ -822,7 +822,11 @@ async function connectWhep() {
 
   pc.ontrack = function(event) {
     log('remote track: ' + event.track.kind + ' (' + Math.round(performance.now() - t0) + 'ms)', 'ok');
-    document.getElementById('outputVideo').srcObject = event.streams[0];
+    var video = document.getElementById('outputVideo');
+    video.srcObject = event.streams[0];
+    video.play().then(function() {
+      video.muted = false;
+    }).catch(function(e) { log('play error: ' + e.message, 'error'); });
   };
 
   pc.addTransceiver('video', { direction: 'recvonly' });