changes for vp9 support

facefusion/apis/stream_video.py

@@ -12,7 +12,7 @@ import numpy
 from facefusion import rtc, state_manager, streamer
 from facefusion.apis.stream_event import create_receive_event
 from facefusion.audio import create_empty_audio_frame
-from facefusion.codecs import aom_decoder, aom_encoder, vp9_decoder, vp9_encoder, vpx_decoder, vpx_encoder
+from facefusion.codecs import aom_decoder, aom_encoder, vpx_decoder, vpx_encoder
 from facefusion.types import AomDecoder, AomEncoder, AomPointer, BitRate, Resolution, RtcPeer, RtcPeerVideo, VideoCodec, VideoPack, VisionFrame, VpxDecoder, VpxEncoder, VpxPointer
 
 
@@ -118,18 +118,12 @@ def decode_video_frame(video_codec : VideoCodec, video_decoder : VpxDecoder | Ao
 		if aom_pointer:
 			return normalize_vision_frame(aom_pointer)
 
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		vpx_pointer = vpx_decoder.decode(video_decoder, input_buffer)
 
 		if vpx_pointer:
 			return normalize_vision_frame(vpx_pointer)
 
-	if video_codec == 'vp9':
-		vpx_pointer = vp9_decoder.decode(video_decoder, input_buffer)
-
-		if vpx_pointer:
-			return normalize_vision_frame(vpx_pointer)
-
 	return None
 
 
@@ -137,12 +131,9 @@ def encode_video_frame(video_codec : VideoCodec, video_encoder : VpxEncoder | Ao
 	if video_codec == 'av1':
 		return aom_encoder.encode(video_encoder, input_buffer, frame_resolution, frame_index)
 
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		return vpx_encoder.encode(video_encoder, input_buffer, frame_resolution, frame_index)
 
-	if video_codec == 'vp9':
-		return vp9_encoder.encode(video_encoder, input_buffer, frame_resolution, frame_index)
-
 	return bytes()
 
 
@@ -156,11 +147,8 @@ def create_video_decoder(video_codec : VideoCodec) -> Optional[VpxDecoder | AomD
 	if video_codec == 'av1':
 		return aom_decoder.create(8)
 
-	if video_codec == 'vp8':
-		return vpx_decoder.create(8)
-
-	if video_codec == 'vp9':
-		return vp9_decoder.create(8)
+	if video_codec in [ 'vp8', 'vp9' ]:
+		return vpx_decoder.create(video_codec, 8)
 
 	return None
 
@@ -169,11 +157,8 @@ def create_video_encoder(video_codec : VideoCodec, frame_resolution : Resolution
 	if video_codec == 'av1':
 		return aom_encoder.create(frame_resolution, bitrate, 8, 10)
 
-	if video_codec == 'vp8':
-		return vpx_encoder.create(frame_resolution, bitrate, 8, 10)
-
-	if video_codec == 'vp9':
-		return vp9_encoder.create(frame_resolution, bitrate, 8, 10)
+	if video_codec in [ 'vp8', 'vp9' ]:
+		return vpx_encoder.create(video_codec, frame_resolution, bitrate, 8, 10)
 
 	return None
 
@@ -182,34 +167,25 @@ def destroy_video_decoder(video_codec : VideoCodec, video_decoder : VpxDecoder |
 	if video_codec == 'av1':
 		aom_decoder.destroy(video_decoder)
 
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		vpx_decoder.destroy(video_decoder)
 
-	if video_codec == 'vp9':
-		vp9_decoder.destroy(video_decoder)
-
 
 def destroy_video_encoder(video_codec : VideoCodec, video_encoder : VpxEncoder | AomEncoder) -> None:
 	if video_codec == 'av1':
 		aom_encoder.destroy(video_encoder)
 
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		vpx_encoder.destroy(video_encoder)
 
-	if video_codec == 'vp9':
-		vp9_encoder.destroy(video_encoder)
-
 
 def update_video_encoder_bitrate(video_codec : VideoCodec, video_encoder : VpxEncoder | AomEncoder, bitrate : BitRate) -> bool:
 	if video_codec == 'av1':
 		return aom_encoder.update_bitrate(video_encoder, bitrate)
 
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		return vpx_encoder.update_bitrate(video_encoder, bitrate)
 
-	if video_codec == 'vp9':
-		return vp9_encoder.update_bitrate(video_encoder, bitrate)
-
 	return False
 
 

facefusion/codecs/vpx_decoder.py

@@ -3,10 +3,10 @@ import struct
 from typing import Optional
 
 from facefusion.libraries import vpx as vpx_module
-from facefusion.types import VideoCodec, VpxDecoder, VpxPointer
+from facefusion.types import VpxDecoder, VpxPointer, VxpVideoCodec
 
 
-def create(video_codec : VideoCodec, thread_count : int) -> Optional[VpxDecoder]:
+def create(video_codec : VxpVideoCodec, thread_count : int) -> Optional[VpxDecoder]:
 	vpx_library = vpx_module.create_static_library()
 
 	if vpx_library:

facefusion/codecs/vpx_encoder.py

@@ -3,10 +3,10 @@ import struct
 from typing import Optional
 
 from facefusion.libraries import vpx as vpx_module
-from facefusion.types import BitRate, Resolution, VideoCodec, VpxEncoder
+from facefusion.types import BitRate, Resolution, VideoCodec, VpxEncoder, VxpVideoCodec
 
 
-def create(video_codec : VideoCodec, frame_resolution : Resolution, bitrate : BitRate, thread_count : int, cpu_count : int) -> Optional[VpxEncoder]:
+def create(video_codec : VxpVideoCodec, frame_resolution : Resolution, bitrate : BitRate, thread_count : int, cpu_count : int) -> Optional[VpxEncoder]:
 	vpx_library = vpx_module.create_static_library()
 
 	if vpx_library:

facefusion/types.py

@@ -99,6 +99,9 @@ Resolution : TypeAlias = Tuple[int, int]
 AudioCodec : TypeAlias = Literal['opus']
 VideoCodec : TypeAlias = Literal['av1', 'vp8', 'vp9']
 
+AomVideoCodec : TypeAlias = Literal['av1']
+VxpVideoCodec : TypeAlias = Literal['vp8', 'vp9']
+
 FrameHandler : TypeAlias = Callable[..., None]
 
 AomEncoder : TypeAlias = ctypes.Array[ctypes.c_char]

tests/test_api_stream_video.py

@@ -11,7 +11,7 @@ import pytest
 
 from facefusion import rtc, rtc_store, state_manager
 from facefusion.apis.stream_video import create_video_decoder, create_video_encoder, decode_video_frame, destroy_video_decoder, destroy_video_encoder, encode_video_frame, handle_video_frame, receive_video_frames, run_video_encode_loop, update_video_encoder_bitrate
-from facefusion.codecs import aom_encoder, vp9_encoder, vpx_encoder
+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
 from facefusion.hash_helper import create_hash
@@ -165,12 +165,9 @@ def test_create_and_destroy_video_decoder(video_codec : VideoCodec) -> None:
 	if video_codec == 'av1':
 		video_encoder = aom_encoder.create((426, 226), 1000, 1, 0)
 		encode_buffer = aom_encoder.encode(video_encoder, input_buffer, (426, 226), 0)
-	if video_codec == 'vp8':
-		video_encoder = vpx_encoder.create((426, 226), 1000, 1, 0)
+	if video_codec in [ 'vp8', 'vp9' ]:
+		video_encoder = vpx_encoder.create(video_codec, (426, 226), 1000, 1, 0)
 		encode_buffer = vpx_encoder.encode(video_encoder, input_buffer, (426, 226), 0)
-	if video_codec == 'vp9':
-		video_encoder = vp9_encoder.create((426, 226), 1000, 1, 0)
-		encode_buffer = vp9_encoder.encode(video_encoder, input_buffer, (426, 226), 0)
 
 	video_decoder = create_video_decoder(video_codec)
 
@@ -189,19 +186,15 @@ def test_create_and_destroy_video_encoder(video_codec : VideoCodec) -> None:
 
 	if video_codec == 'av1':
 		assert aom_encoder.encode(video_encoder, input_buffer, (426, 226), 0)
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		assert vpx_encoder.encode(video_encoder, input_buffer, (426, 226), 0)
-	if video_codec == 'vp9':
-		assert vp9_encoder.encode(video_encoder, input_buffer, (426, 226), 0)
 
 	destroy_video_encoder(video_codec, video_encoder)
 
 	if video_codec == 'av1':
 		assert aom_encoder.encode(video_encoder, input_buffer, (426, 226), 1) == bytes()
-	if video_codec == 'vp8':
+	if video_codec in [ 'vp8', 'vp9' ]:
 		assert vpx_encoder.encode(video_encoder, input_buffer, (426, 226), 1) == bytes()
-	if video_codec == 'vp9':
-		assert vp9_encoder.encode(video_encoder, input_buffer, (426, 226), 1) == bytes()
 
 
 @pytest.mark.parametrize('video_codec', [ 'av1', 'vp8', 'vp9' ])

tests/test_codec_vpx_decoder.py

@@ -11,6 +11,7 @@ from facefusion.common_helper import is_linux, is_macos, is_windows
 from facefusion.download import conditional_download
 from facefusion.hash_helper import create_hash
 from facefusion.libraries import vpx as vpx_module
+from facefusion.types import VxpVideoCodec
 from facefusion.vision import read_video_frame
 
 
@@ -23,30 +24,41 @@ def before_all() -> None:
 	vpx_module.pre_check()
 
 
-def test_create() -> None:
-	assert create(1)
+@pytest.mark.parametrize('video_codec', [ 'vp8', 'vp9' ])
+def test_create(video_codec : VxpVideoCodec) -> None:
+	assert create(video_codec, 1)
 
 	with patch('facefusion.libraries.vpx.create_static_library', return_value = None):
-		assert create(1) is None
+		assert create(video_codec, 1) is None
 
 
-def test_decode() -> None:
+@pytest.mark.parametrize('video_codec', [ 'vp8', 'vp9' ])
+def test_decode(video_codec : VxpVideoCodec) -> None:
 	video_frame = read_video_frame(get_test_example_file('target-240p.mp4'))
 	video_buffer = cv2.cvtColor(video_frame, cv2.COLOR_BGR2YUV_I420).tobytes()
 	video_resolution = (video_frame.shape[1], video_frame.shape[0])
-	vpx_encoder = create_encoder(video_resolution, 1000, 1, 0)
+	vpx_encoder = create_encoder(video_codec, video_resolution, 1000, 1, 0)
 	encode_buffer = encode(vpx_encoder, video_buffer, video_resolution, 0)
-	vpx_decoder = create(1)
+	vpx_decoder = create(video_codec, 1)
 
 	if is_linux() or is_windows():
-		assert create_hash(decode(vpx_decoder, encode_buffer).get('buffer')) == 'dc9c8864'
+		if video_codec == 'vp8':
+			assert create_hash(decode(vpx_decoder, encode_buffer).get('buffer')) == 'dc9c8864'
+
+		if video_codec == 'vp9':
+			assert create_hash(decode(vpx_decoder, encode_buffer).get('buffer')) == '50f4c1fa'
 
 	if is_macos():
-		assert create_hash(decode(vpx_decoder, encode_buffer).get('buffer')) == '87450f70'
+		if video_codec == 'vp8':
+			assert create_hash(decode(vpx_decoder, encode_buffer).get('buffer')) == '87450f70'
+
+		if video_codec == 'vp9':
+			pytest.skip()
 
 
-def test_destroy() -> None:
-	vpx_decoder = create(1)
+@pytest.mark.parametrize('video_codec', [ 'vp8', 'vp9' ])
+def test_destroy(video_codec : VxpVideoCodec) -> None:
+	vpx_decoder = create(video_codec, 1)
 
 	with patch.object(vpx_module.create_static_library(), 'vpx_codec_destroy') as mock:
 		destroy(vpx_decoder)

tests/test_codec_vpx_encoder.py

@@ -2,6 +2,8 @@ from unittest.mock import patch
 
 import cv2
 import pytest
+
+from facefusion.types import VxpVideoCodec
 from tests.assert_helper import get_test_example_file, get_test_examples_directory
 
 from facefusion import state_manager
@@ -22,28 +24,35 @@ def before_all() -> None:
 	vpx_module.pre_check()
 
 
-def test_create() -> None:
-	assert create((320, 240), 1000, 8, 16)
+@pytest.mark.parametrize('video_codec', [ 'vp8', 'vp9' ])
+def test_create(video_codec : VxpVideoCodec) -> None:
+	assert create(video_codec, (320, 240), 1000, 8, 16)
 
 	with patch('facefusion.libraries.vpx.create_static_library', return_value = None):
-		assert create((320, 240), 1000, 8, 16) is None
+		assert create(video_codec, (320, 240), 1000, 8, 16) is None
 
 
-def test_encode() -> None:
+@pytest.mark.parametrize('video_codec', [ 'vp8', 'vp9' ])
+def test_encode(video_codec : VxpVideoCodec) -> None:
 	video_frame = read_video_frame(get_test_example_file('target-240p.mp4'))
 	video_buffer = cv2.cvtColor(video_frame, cv2.COLOR_BGR2YUV_I420).tobytes()
 	video_resolution = (video_frame.shape[1], video_frame.shape[0])
-	vpx_encoder = create(video_resolution, 1000, 1, 0)
+	vpx_encoder = create(video_codec, video_resolution, 1000, 1, 0)
 
 	if is_linux() or is_windows():
-		assert create_hash(encode(vpx_encoder, video_buffer, video_resolution, 3)) == 'ce133a1f'
+		if video_codec == 'vp8':
+			assert create_hash(encode(vpx_encoder, video_buffer, video_resolution, 3)) == 'ce133a1f'
+
+		if video_codec == 'vp9':
+			assert create_hash(encode(vpx_encoder, video_buffer, video_resolution, 3)) == 'b0760d5e'
 
 	if is_macos():
 		pytest.skip()
 
 
-def test_destroy() -> None:
-	vpx_encoder = create((320, 240), 1000, 8, 16)
+@pytest.mark.parametrize('video_codec', [ 'vp8', 'vp9' ])
+def test_destroy(video_codec : VxpVideoCodec) -> None:
+	vpx_encoder = create(video_codec, (320, 240), 1000, 8, 16)
 
 	with patch.object(vpx_module.create_static_library(), 'vpx_codec_destroy') as mock:
 		destroy(vpx_encoder)