basic cpu encode support

This commit is contained in:
Jerry Yan 2024-11-27 16:47:39 +08:00
parent a8abb92b84
commit 9139727dc6
3 changed files with 136 additions and 23 deletions

View File

@ -1,3 +1,5 @@
import os.path
from entity.ffmpeg import FfmpegTask
import logging
@ -10,32 +12,32 @@ def parse_ffmpeg_task(task_info, template_info):
tasks = []
# 中间片段
for part in template_info.get("video_parts"):
sub_ffmpeg_task = parse_video_part(part, task_info)
if not sub_ffmpeg_task:
source = select_video_if_needed(part.get('source'), task_info, template_info)
if not source:
logger.warning("no video found for part: " + str(part))
continue
sub_ffmpeg_task = FfmpegTask(source)
sub_ffmpeg_task.frame_rate = template_info.get("frame_rate", 25)
for lut in part.get('filters', []):
sub_ffmpeg_task.add_lut(os.path.join(template_info.get("local_path"), lut))
for audio in part.get('audios', []):
sub_ffmpeg_task.add_audios(os.path.join(template_info.get("local_path"), audio))
for overlay in part.get('overlays', []):
sub_ffmpeg_task.add_overlay(os.path.join(template_info.get("local_path"), overlay))
tasks.append(sub_ffmpeg_task)
task = FfmpegTask(tasks, output_file="test.mp4")
task.correct_task_type()
overall = template_info.get("overall_template")
task.add_lut(*overall.get('luts', []))
task.add_audios(*overall.get('audios', []))
task.add_overlay(*overall.get('overlays', []))
task.frame_rate = template_info.get("frame_rate", 25)
for lut in overall.get('filters', []):
task.add_lut(os.path.join(template_info.get("local_path"), lut))
for audio in overall.get('audios', []):
task.add_audios(os.path.join(template_info.get("local_path"), audio))
for overlay in overall.get('overlays', []):
task.add_overlay(os.path.join(template_info.get("local_path"), overlay))
return task
def parse_video_part(video_part, task_info):
source = select_video_if_needed(video_part.get('source'), task_info)
if not source:
logger.warning("no video found for part: " + str(video_part))
return None
task = FfmpegTask(source)
task.add_lut(*video_part.get('luts', []))
task.add_audios(*video_part.get('audios', []))
task.add_overlay(*video_part.get('overlays', []))
return task
def select_video_if_needed(source, task_info):
def select_video_if_needed(source, task_info, template_info):
if source.startswith('PLACEHOLDER_'):
placeholder_id = source.replace('PLACEHOLDER_', '')
new_sources = task_info.get('user_videos', {}).get(placeholder_id, [])
@ -47,7 +49,7 @@ def select_video_if_needed(source, task_info):
# TODO: Random Pick / Policy Pick
new_sources = new_sources[0]
return new_sources
return source
return os.path.join(template_info.get("local_path"), source)
def start_ffmpeg_task(ffmpeg_task):

View File

@ -12,6 +12,7 @@ class FfmpegTask(object):
self.output_file = output_file
self.mute = True
self.speed = 1
self.frame_rate = 25
self.subtitles = []
self.luts = []
self.audios = []
@ -36,7 +37,7 @@ class FfmpegTask(object):
for i in self.input_file:
if type(i) is str:
continue
elif type(i) is FfmpegTask:
elif isinstance(i, FfmpegTask):
if i.need_run():
yield i
@ -69,7 +70,9 @@ class FfmpegTask(object):
def get_output_file(self):
if self.task_type == 'copy':
return self.input_file
return self.input_file[0]
if self.output_file == '':
self.set_output_file()
return self.output_file
def correct_task_type(self):
@ -109,3 +112,103 @@ class FfmpegTask(object):
def check_audio_track(self):
if len(self.audios) > 0:
self.mute = False
def get_ffmpeg_args(self):
args = ['-y', '-hide_banner']
video_output_str = "[0:v]"
if self.task_type == 'encode':
input_args = []
filter_args = []
output_args = []
video_output_str = "[0:v]"
audio_output_str = "[0:v]"
video_input_count = 0
audio_input_count = 0
for input_file in self.input_file:
input_args.append("-i")
if type(input_file) is str:
input_args.append(input_file)
elif isinstance(input_file, FfmpegTask):
input_args.append(input_file.get_output_file())
for lut in self.luts:
filter_args.append("[0:v]lut3d=file=" + lut + "[0:v]")
for overlay in self.overlays:
input_index = input_args.count("-i")
input_args.append("-i")
input_args.append(overlay)
filter_args.append(f"{video_output_str}[{input_index}:v]scale=rw:rh[v]")
filter_args.append(f"[v][{input_index}:v]overlay=1:eof_action=endall[v]")
video_output_str = "[v]"
output_args.append("-map")
output_args.append(video_output_str)
output_args.append("-r")
output_args.append(f"{self.frame_rate}")
if self.mute:
output_args.append("-an")
else:
input_index = 0
for audio in self.audios:
input_index = input_args.count("-i")
input_args.append("-i")
input_args.append(audio.replace("\\", "/"))
if audio_input_count > 0:
filter_args.append(f"{audio_output_str}[{input_index}:a]amix[a]")
audio_output_str = "[a]"
else:
audio_output_str = f"[{input_index}:a]"
audio_input_count += 1
if audio_input_count == 1:
audio_output_str = f"{input_index}"
output_args.append(f"-map")
output_args.append(audio_output_str)
return args + input_args + ["-filter_complex", ";".join(filter_args)] + output_args + [self.get_output_file()]
elif self.task_type == 'concat':
input_args = []
output_args = []
filter_args = []
video_output_str = "[0:v]"
audio_output_str = "[0:v]"
video_input_count = 0
audio_input_count = 0
for input_file in self.input_file:
input_index = input_args.count("-i")
input_args.append("-i")
if type(input_file) is str:
input_args.append(input_file.replace("\\", "/"))
elif isinstance(input_file, FfmpegTask):
input_args.append(input_file.get_output_file().replace("\\", "/"))
if video_input_count > 0:
filter_args.append(f"{video_output_str}[{input_index}:v]concat=n=2:v=1:a=0[v]")
video_output_str = "[v]"
else:
video_output_str = f"[{input_index}:v]"
video_input_count += 1
output_args.append("-map")
output_args.append(video_output_str)
if self.mute:
output_args.append("-an")
else:
input_index = 0
for audio in self.audios:
input_index = input_args.count("-i")
input_args.append("-i")
input_args.append(audio.replace("\\", "/"))
if audio_input_count > 0:
filter_args.append(f"{audio_output_str}[{input_index}:a]amix[a]")
audio_output_str = "[a]"
else:
audio_output_str = f"[{input_index}:a]"
audio_input_count += 1
if audio_input_count == 1:
audio_output_str = f"{input_index}"
output_args.append(f"-map")
output_args.append(audio_output_str)
return args + input_args + ["-filter_complex", ";".join(filter_args)] + output_args + [self.get_output_file()]
def set_output_file(self, file=None):
if file is None:
if self.output_file == '':
# TODO: Random Filename
self.output_file = "rand.mp4"
else:
self.output_file = file

View File

@ -1,3 +1,5 @@
import os
from datetime import datetime
from typing import Optional, IO
from entity.ffmpeg import FfmpegTask
@ -5,6 +7,8 @@ from entity.ffmpeg import FfmpegTask
def start_render(ffmpeg_task: FfmpegTask):
print(ffmpeg_task)
print(ffmpeg_task.get_ffmpeg_args())
os.system("ffmpeg.exe "+" ".join(ffmpeg_task.get_ffmpeg_args()))
def handle_ffmpeg_output(stdout: Optional[IO[bytes]]) -> str:
out_time = "0:0:0.0"
@ -24,4 +28,8 @@ def handle_ffmpeg_output(stdout: Optional[IO[bytes]]) -> str:
if line.startswith(b"speed="):
speed = line.replace(b"speed=", b"").decode().strip()
print("[ ]Speed:", out_time, "@", speed)
return out_time
return out_time
def duration_str_to_float(duration_str: str) -> float:
_duration = datetime.strptime(duration_str, "%H:%M:%S.%f") - datetime(1900, 1, 1)
return _duration.total_seconds()