CalvinBackup/facefusion
1
0
Fork 0
mirror of https://github.com/facefusion/facefusion.git synced 2026-07-25 11:30:54 +02:00
Code Issues Packages Projects Releases Wiki Activity

Refactor(stream-helper): split encode/receive loops and unify audio/video structure (#1142)

Browse Source 
* split encode loop

* unify audio and video encode loop methods

* improve variable names

* split receive methods

* test improve

* try to avoid != with even more weird approach

* remove source_path check

* remove empty variables

* fix lint

* avoid not in condition

* rename

* fix lint
This commit is contained in:
Harisreedhar
2026-06-02 01:50:40 +05:30
committed by GitHub
parent e2e4e6a95b
commit beeb1d99e9
2 changed files with 172 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test_api_stream_helper.py
+74 -47
View File
@@ -10,7 +10,7 @@ import numpy
import pytest
from facefusion import rtc, rtc_store, state_manager
from facefusion.apis.stream_helper import create_video_decoder, create_video_encoder, decode_video_frame, destroy_stream, destroy_video_decoder, destroy_video_encoder, encode_video_frame, process_image, process_video, receive_audio_frames, receive_video_frames, receive_vision_frames, run_encode_loop, run_peer_loop, update_video_encoder_bitrate
from facefusion.apis.stream_helper import create_video_decoder, create_video_encoder, decode_video_frame, destroy_stream, destroy_video_decoder, destroy_video_encoder, encode_video_frame, fill_audio_deque, fill_video_deque, process_image, process_video, receive_audio_frames, receive_video_frames, receive_vision_frames, run_audio_encode_loop, run_peer_loop, run_video_encode_loop, update_video_encoder_bitrate
from facefusion.codecs import aom_encoder, vpx_encoder
from facefusion.common_helper import is_linux, is_macos, is_windows
from facefusion.download import conditional_download
@@ -57,7 +57,6 @@ async def test_process_image() -> None:
}
]
state_manager.init_item('source_paths', [ get_test_example_file('source.jpg') ])
await process_image(websocket_mock)
websocket_mock.send_bytes.assert_called_once()
@@ -65,11 +64,6 @@ async def test_process_image() -> None:
if is_linux() or is_windows():
assert create_hash(websocket_mock.send_bytes.call_args[0][0]) == '0142782f'
state_manager.init_item('source_paths', None)
await process_image(websocket_mock)
websocket_mock.send_bytes.assert_called_once()
@pytest.mark.parametrize('video_codec, session_id', [ ('av1', 'test-process-video-av1'), ('vp8', 'test-process-video-vp8') ])
def test_process_video(video_codec : VideoCodec, session_id : str) -> None:
@@ -155,7 +149,7 @@ def test_run_peer_loop(video_codec : VideoCodec, payload_type : int, session_id
assert rtc_store.has_peers(session_id) is True
with patch('facefusion.apis.stream_helper.receive_video_frames'):
with patch('facefusion.apis.stream_helper.run_encode_loop'):
with patch('facefusion.apis.stream_helper.run_video_encode_loop'):
thread = threading.Thread(target = asyncio.run, args = (run_peer_loop(session_id, rtc_peer),), daemon = True)
thread.start()
thread.join(timeout = 5.0)
@@ -164,7 +158,7 @@ def test_run_peer_loop(video_codec : VideoCodec, payload_type : int, session_id
@pytest.mark.parametrize('video_codec, payload_type', [ ('av1', 35), ('vp8', 96) ])
def test_run_encode_loop(video_codec : VideoCodec, payload_type : int) -> None:
def test_run_video_encode_loop(video_codec : VideoCodec, payload_type : int) -> None:
video_frame = read_video_frame(get_test_example_file('target-240p.mp4'))
peer_connection = rtc.create_peer_connection()
video_sender_track = rtc.add_video_track(peer_connection, 'sendonly', video_codec, payload_type)
@@ -183,14 +177,13 @@ def test_run_encode_loop(video_codec : VideoCodec, payload_type : int) -> None:
}
video_deque : deque[VideoPack] = deque()
audio_deque : deque[AudioPack] = deque()
video_event = threading.Event()
video_deque.append((video_frame, 0.1))
video_event.set()
with patch('facefusion.apis.stream_helper.rtc.send_video') as send_video_mock:
encode_loop_thread = threading.Thread(target = run_encode_loop, args = (rtc_peer, video_codec, video_deque, audio_deque, video_event), daemon = True)
encode_loop_thread = threading.Thread(target = run_video_encode_loop, args = (rtc_peer, video_deque, video_event), daemon = True)
encode_loop_thread.start()
empty_vision_frame = numpy.empty(0)
video_deque.append((empty_vision_frame, 0.0))
@@ -210,60 +203,73 @@ def test_run_encode_loop(video_codec : VideoCodec, payload_type : int) -> None:
pytest.skip()
@pytest.mark.parametrize('video_codec, payload_type', [ ('av1', 35), ('vp8', 96) ])
def test_run_peer_loop_send_order(video_codec : VideoCodec, payload_type : int) -> None:
video_frame = read_video_frame(get_test_example_file('target-240p.mp4'))
def test_run_audio_encode_loop() -> None:
audio_buffer = read_audio_buffer(get_test_example_file('source.mp3'), 48000, 16, 2)
audio_frame = numpy.frombuffer(audio_buffer, dtype = numpy.int16).astype(numpy.float32) / 32768.0
peer_connection = rtc.create_peer_connection()
video_sender_track = rtc.add_video_track(peer_connection, 'sendonly', video_codec, payload_type)
video_receiver_track = rtc.add_video_track(peer_connection, 'recvonly', video_codec, payload_type)
rtc_peer : RtcPeer =\
{
'peer_connection': peer_connection,
'video':
{
'sender_track': video_sender_track,
'receiver_track': video_receiver_track,
'codec': video_codec
'sender_track': 0,
'receiver_track': 0,
'codec': 'vp8'
},
'sender_bitrate': ctypes.c_uint(0),
'receiver_bitrate': ctypes.c_uint(0)
}
video_deque : deque[VideoPack] = deque()
audio_deque : deque[AudioPack] = deque()
video_event = threading.Event()
audio_event = threading.Event()
video_deque.append((video_frame, 0.100))
audio_deque.append((audio_frame, 0.100))
video_event.set()
streamer_mock = MagicMock()
streamer_mock.process_frame.return_value = video_frame
audio_event.set()
encoder_mock = MagicMock()
encoder_mock.encode.return_value = bytes([ 1 ] * 32)
rtc_mock = MagicMock()
with patch('facefusion.apis.stream_helper.streamer.process_frame', streamer_mock.process_frame):
with patch('facefusion.apis.stream_helper.opus_encoder.encode', encoder_mock.encode):
with patch('facefusion.apis.stream_helper.rtc.send_audio', rtc_mock.send_audio):
with patch('facefusion.apis.stream_helper.rtc.send_video', rtc_mock.send_video):
encode_loop_thread = threading.Thread(target = run_encode_loop, args = (rtc_peer, video_codec, video_deque, audio_deque, video_event), daemon = True)
encode_loop_thread.start()
empty_vision_frame = numpy.empty(0)
video_deque.append((empty_vision_frame, 0.0))
video_event.set()
encode_loop_thread.join(timeout = 5.0)
assert streamer_mock.process_frame.called is True
with patch('facefusion.apis.stream_helper.opus_encoder.encode', encoder_mock.encode):
with patch('facefusion.apis.stream_helper.rtc.send_audio') as send_audio_mock:
audio_loop_thread = threading.Thread(target = run_audio_encode_loop, args = (rtc_peer, audio_deque, audio_event), daemon = True)
audio_loop_thread.start()
audio_deque.append((numpy.empty(0), 0.0))
audio_event.set()
audio_loop_thread.join(timeout = 5.0)
assert encoder_mock.encode.called is True
assert send_audio_mock.called is True
assert rtc_mock.send_audio.called is True
assert rtc_mock.send_video.called is True
@pytest.mark.parametrize('video_codec', [ 'av1', 'vp8' ])
def test_fill_video_deque(video_codec : VideoCodec) -> None:
video_frame = read_video_frame(get_test_example_file('target-240p.mp4'))
input_buffer = cv2.cvtColor(video_frame, cv2.COLOR_BGR2YUV_I420).tobytes()
video_encoder = create_video_encoder(video_codec, (426, 226), 1000)
video_decoder = create_video_decoder(video_codec)
encode_buffer = encode_video_frame(video_codec, video_encoder, input_buffer, (426, 226), 0)
video_deque : deque[VideoPack] = deque()
video_event = threading.Event()
fill_video_deque(video_codec, video_decoder, encode_buffer, video_deque, video_event)
vision_frame, _ = video_deque.popleft()
assert video_event.is_set()
if is_linux() or is_windows():
if video_codec == 'av1':
assert create_hash(vision_frame.tobytes()) == 'c97d6d29'
if video_codec == 'vp8':
assert create_hash(vision_frame.tobytes()) == '99ef2c25'
if is_macos():
if video_codec == 'av1':
assert create_hash(vision_frame.tobytes()) == 'eafd1fab'
if video_codec == 'vp8':
assert create_hash(vision_frame.tobytes()) == 'ff3ecb43'
@pytest.mark.parametrize('video_codec', [ 'av1', 'vp8' ])
@@ -287,11 +293,29 @@ def test_receive_video_frames(video_codec : VideoCodec) -> None:
video_receiver_thread.start()
video_receiver_thread.join(timeout = 5.0)
vision_frame, _ = video_deque.popleft()
if is_linux() or is_windows():
assert create_hash(video_deque[0][0].tobytes()) == 'a17439db'
assert create_hash(vision_frame.tobytes()) == 'a17439db'
if is_macos():
assert create_hash(video_deque[0][0].tobytes()) == '38d00e2a'
assert create_hash(vision_frame.tobytes()) == '38d00e2a'
def test_fill_audio_deque() -> None:
audio_buffer = read_audio_buffer(get_test_example_file('source.mp3'), 48000, 16, 2)
audio_frame = numpy.frombuffer(audio_buffer, dtype = numpy.int16).astype(numpy.float32) / 32768.0
audio_decoder_mock = MagicMock()
audio_deque : deque[AudioPack] = deque()
audio_event = threading.Event()
with patch('facefusion.apis.stream_helper.decode_audio_frame', return_value = audio_frame.tobytes()):
fill_audio_deque('opus', audio_decoder_mock, audio_frame.tobytes(), audio_deque, audio_event)
buffer_frame, _ = audio_deque.popleft()
assert audio_event.is_set()
assert create_hash(buffer_frame.tobytes()) == create_hash(audio_frame.tobytes())
@pytest.mark.parametrize('audio_codec', [ 'opus' ])
@@ -299,23 +323,26 @@ def test_receive_audio_frames(audio_codec : AudioCodec) -> None:
audio_buffer = read_audio_buffer(get_test_example_file('source.mp3'), 48000, 16, 2)
audio_frame = numpy.frombuffer(audio_buffer, dtype = numpy.int16).astype(numpy.float32) / 32768.0
audio_deque : deque[AudioPack] = deque()
audio_event = threading.Event()
datachannel_library_mock = MagicMock()
datachannel_library_mock.rtcReceiveMessage.side_effect = [ 0, -1 ]
with patch('facefusion.apis.stream_helper.datachannel_module.create_static_library', return_value = datachannel_library_mock):
with patch('facefusion.apis.stream_helper.opus_decoder.decode', return_value = audio_frame.tobytes()):
with patch('facefusion.apis.stream_helper.decode_audio_frame', return_value = audio_frame.tobytes()):
rtc_peer_audio : RtcPeerAudio =\
{
'sender_track': 0,
'receiver_track': 0,
'codec': audio_codec
}
audio_receiver_thread = threading.Thread(target = receive_audio_frames, args = (rtc_peer_audio, audio_deque), daemon = True)
audio_receiver_thread = threading.Thread(target = receive_audio_frames, args = (rtc_peer_audio, audio_deque, audio_event), daemon = True)
audio_receiver_thread.start()
audio_receiver_thread.join(timeout = 5.0)
assert create_hash(audio_deque[0][0].tobytes()) == create_hash(audio_frame.tobytes())
buffer_frame, _ = audio_deque.popleft()
assert create_hash(buffer_frame.tobytes()) == create_hash(audio_frame.tobytes())
@pytest.mark.parametrize('video_codec', [ 'av1', 'vp8' ])