nodes/MoonHugo-ComfyUI-FFmpeg
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

修改缝合视频ffmpeg功能，支持裁剪和背景填充；添加画中画ffmpeg节点。

Browse Source
This commit is contained in:
sunjinshu
2025-05-25 20:05:18 +08:00
parent 89c52e1789
commit 42d8e25257
2 changed files with 206 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
nodes/pipVideo.py
+166
View File
@@ -0,0 +1,166 @@
import os
import subprocess
from ..func import has_audio,getVideoInfo,set_file_name,video_type
import torch
import math
device = "cuda" if torch.cuda.is_available() else "cpu"
class PipVideo:
def __init__(self):
pass
@classmethod
def INPUT_TYPES(cls):
return {
"required": {
"video1_path": ("STRING", {"default":"C:/Users/Desktop/video1.mp4", "tooltip": "说明：画中画背景画面！"}),
"video2_path": ("STRING", {"default":"C:/Users/Desktop/video2.mp4", "tooltip": "说明：画中画前景画面！"}),
"device": (["cpu","cuda"], {"default":device,}),
"use_audio": (["video1","video2"], {"default":"video1", "tooltip": "说明：最终视频使用哪个视频的音轨！"}),
"use_duration": (["video1","video2"], {"default":"video2", "tooltip": "说明：使用哪个视频作为最终参考时长！"}),
"align_type":(["top-left","top-right", "bottom-left", "bottom-right", "center"], {"default":"center",}),
"pip_fg_zoom": ("FLOAT", { "default": 2.5, "min": 1, "max": 100, "step": 0.5, "tooltip": "说明：画中画背景缩放系数，越大前景画面越小，值为背景宽高的缩小倍数！"}),
"output_path": ("STRING", {"default": "C:/Users/Desktop/output"}),
"scale_and_crop": (["none","540*960", "960*540"], {"default": "none", "tooltip": "说明：缩放和裁剪比例！"}), # 新增参数控制缩放裁剪
"fps": ("FLOAT", {"min": 0, "max": 60, "step": 0.1, "default": 30.0, "tooltip": "说明：画中画合并后的强制帧率，设置为0将使用video2的帧率，设置为1为将使用video2的帧率！"}),
"is_chromakey": ("BOOLEAN", { "default": False ,"label_on": "绿幕去背景", "label_off": "关闭绿幕透明", "tooltip": "说明：是否进行绿幕去背景！"}), #是否画中画 绿幕透明
},
}
RETURN_TYPES = ("STRING","INT","INT","FLOAT","FLOAT",)
RETURN_NAMES = ("video_complete_paths","width","height","duration","fps",)
FUNCTION = "pip_video"
OUTPUT_NODE = True
CATEGORY = "🔥FFmpeg"
DESCRIPTION = """两个视频叠加成一个画中画效果，可以控制前景video2出现在前景video1画面上的位置，
可以设置前景画面的缩放系数和是否去掉绿幕背景等设置."""
def pip_video(self, video1_path, video2_path,device,use_audio,use_duration, align_type,pip_fg_zoom, output_path,scale_and_crop,fps,is_chromakey):
try:
video1_path = os.path.abspath(video1_path).strip()
video2_path = os.path.abspath(video2_path).strip()
output_path = os.path.abspath(output_path).strip()
# 视频不存在
if not video1_path.lower().endswith(video_type()):
raise ValueError("video1_path："+video1_path+"不是视频文件（video1_path:"+video1_path+" is not a video file）")
if not os.path.isfile(video1_path):
raise ValueError("video1_path："+video1_path+"不存在（video1_path:"+video1_path+" does not exist）")
if not video2_path.lower().endswith(video_type()):
raise ValueError("video2_path："+video2_path+"不是视频文件（video2_path:"+video2_path+" is not a video file）")
if not os.path.isfile(video2_path):
raise ValueError("video2_path："+video2_path+"不存在（video2_path:"+video2_path+" does not exist）")
#判断output_path是否是一个目录
if not os.path.isdir(output_path):
raise ValueError("output_path："+output_path+"不是目录（output_path:"+output_path+" is not a directory）")
video1_audio = has_audio(video1_path)
video2_audio = has_audio(video2_path)
#final_output = set_file_name(video1_path)
#增加以视频1+视频2的文件名合并重命名
final_output = f"{os.path.splitext(os.path.basename(video1_path))[0]}--{os.path.splitext(os.path.basename(video2_path))[0]}.mp4"
#文件名根据年月日时分秒来命名
output_path = os.path.join(output_path, final_output)
use_cuvid = ""
use_encoder = "-c:v libx264" #默认用CPU编码
if device == "cuda":
use_cuvid = "-hwaccel cuda"
use_encoder = "-c:v h264_nvenc"
video_info = getVideoInfo(video1_path)
video_info1 = getVideoInfo(video2_path)
if use_duration == "video1":
duration_1 = video_info['duration']
else:
duration_1 = video_info1['duration']
if fps==0:
fps=video_info['fps']
elif fps==1:
fps=video_info1['fps']
# libx264、libx265等编码器要求宽高必须是2的倍数，如果报错，可以把device换为GPU
width = math.ceil(video_info['width']/2)*2
height = math.ceil(video_info['height']/2)*2
use_audio = {
'video1': '0',
'video2': '1',
}.get(use_audio, '0')
align_position = {
"top-left": f"0:0",
"top-right": f"(W-w):0",
"bottom-left": f"0:(H-h)",
"bottom-right": f"(W-w):(H-h)",
"center": f"(W-w)/2:(H-h)/2",
}.get(align_type, f"(W-w)/2:(H-h)/2")
if height*540/width>=960: #如果高同比缩放高度超出960，需要裁剪
pad_or_crop1='crop=540:960:(iw-iw)/2:(ih-ih/2)/2'
else: #比固定高小，需要填充黑边
pad_or_crop1='pad=540:960:(ow-iw)/2:(oh-ih)/2:color=black'
if height*960/width>=540: #如果缩放到宽960时，高超出540，需要裁剪
pad_or_crop2='crop=960:540:(iw-iw)/2:(ih-ih/2)/2'
else:
pad_or_crop2='pad=960:540:(ow-iw)/2:(oh-ih)/2:color=black'
scale_and_crop_data = {
'none': 'null',
'540*960': f'scale=540:-1,setsar=1,{pad_or_crop1}',
'960*540': f'scale=960:-1,setsar=1,{pad_or_crop2}',
}.get(scale_and_crop, 'null')
video2_width = {
'none': f'{width}',
'540*960': '540',
'960*540': '960',
}.get(scale_and_crop, f'{width}')
final_out = {
'none': f'scale={width}:{height}:force_original_aspect_ratio=disable,setsar=1',
'540*960': 'scale=540:960:force_original_aspect_ratio=disable,setsar=1',
'960*540': 'scale=960:540:force_original_aspect_ratio=disable,setsar=1',
}.get(scale_and_crop, f'scale={width}:{height}:force_original_aspect_ratio=disable,setsar=1')
#测试去绿幕效果
if is_chromakey:
chromakey="chromakey=0x00FF00:0.3:0.1,format=yuva420p"
else:
chromakey="null"
if video1_audio or video2_audio:
#-map 1:a 指定使用第二个视频的音频流
command = fr'ffmpeg "-y" {use_cuvid} -stream_loop -1 -i "{video1_path}" -stream_loop -1 -i "{video2_path}" -filter_complex "[0:v]fps={fps},setpts=PTS-STARTPTS[bg];[1:v]fps={fps},setpts=PTS-STARTPTS[fg];[bg]{scale_and_crop_data}[bg_out];[fg]{chromakey}[fgd];[fgd]scale={video2_width}/{pip_fg_zoom}:-1,setsar=1[fg_out];[bg_out][fg_out]overlay={align_position}[out];[out]{final_out}[final_out]" -map "[final_out]" -map {use_audio}:a? {use_encoder} -c:a aac -t {duration_1} "{output_path}"'
else:
command = fr'ffmpeg "-y" {use_cuvid} -stream_loop -1 -i "{video1_path}" -stream_loop -1 -i "{video2_path}" -filter_complex "[0:v]fps={fps},setpts=PTS-STARTPTS[bg];[1:v]fps={fps},setpts=PTS-STARTPTS[fg];[bg]{scale_and_crop_data}[bg_out];[fg]{chromakey}[fgd];[fgd]scale={video2_width}/{pip_fg_zoom}:-1,setsar=1[fg_out];[bg_out][fg_out]overlay={align_position}[out];[out]{final_out}[final_out]" -map "[final_out]" -t {duration_1} "{output_path}"'
print(f">>>{command}")
# 执行命令并检查错误
result = subprocess.run(command, stderr=subprocess.PIPE, stdout=subprocess.PIPE)
#print("command result",result.returncode)
# 检查返回码
if result.returncode != 0:
# 如果有错误，输出错误信息
print(f"Error: {result.stderr.decode('utf-8')}")
if device == "cuda":
print(f"***当前运算模式*[{device}]*************看下换成CPU重新执行，是否解决因编码问题的报错！********")
self.pip_video(video1_path, video2_path,"cpu",use_audio,use_duration, align_type,pip_fg_zoom, os.path.dirname(output_path),scale_and_crop,fps,is_chromakey)
else:
# 输出标准输出信息
print(f">>FFmpeg 执行完毕！Completed!\t stdout: {result.stdout}")
return (output_path,width,height,duration_1,fps,)
except Exception as e:
raise ValueError(e)
#a=StitchingVideo()
#a.stitching_video("C:/Users/wtc/Desktop/tt/l.mp4", "C:/Users/wtc/Desktop/tt/r.mp4", "cpu", "video2", "horizontal","picture-picture", "bottom-left", "C:/Users/wtc/Desktop/tt", "yes")
nodes/stitchingVideo.py
+40 -9
View File
@@ -3,6 +3,7 @@ import subprocess
from ..func import has_audio,getVideoInfo,set_file_name,video_type
import torch
import math
import time
device = "cuda" if torch.cuda.is_available() else "cpu"
@@ -20,6 +21,7 @@ class StitchingVideo:
"use_audio": (["video1","video2"], {"default":"video1",}),
"stitching_type":(["horizontal","vertical"], {"default":"horizontal",}),
"output_path": ("STRING", {"default": "C:/Users/Desktop/output"}),
"scale_and_crop": (["yes", "no"], {"default": "no"}), # 新增参数控制缩放裁剪
},
}
@@ -29,7 +31,7 @@ class StitchingVideo:
OUTPUT_NODE = True
CATEGORY = "🔥FFmpeg"
def stitching_video(self, video1_path, video2_path,device,use_audio,stitching_type,output_path):
def stitching_video(self, video1_path, video2_path,device,use_audio,stitching_type,output_path,scale_and_crop):
try:
video1_path = os.path.abspath(video1_path).strip()
video2_path = os.path.abspath(video2_path).strip()
@@ -52,10 +54,14 @@ class StitchingVideo:
video1_audio = has_audio(video1_path)
video2_audio = has_audio(video2_path)
final_output = set_file_name(video1_path)
#final_output = set_file_name(video1_path)
#增加以视频1+视频2的文件名合并重命名
final_output = f"{os.path.splitext(os.path.basename(video1_path))[0]}--{os.path.splitext(os.path.basename(video2_path))[0]}.mp4"
#文件名根据年月日时分秒来命名
output_path = os.path.join(output_path, final_output)
#CPU默认参数：
use_cuvid = ""
use_encoder = "-c:v libx264" #默认用CPU编码
@@ -64,7 +70,11 @@ class StitchingVideo:
use_encoder = "-c:v h264_nvenc"
video_info = getVideoInfo(video1_path)
video_info1 = getVideoInfo(video2_path)
duration = video_info['duration']
fps = video_info['fps']
loop_count = max(1, int(duration / video_info1['duration'] + 0.9999))
# libx264、libx265等编码器要求宽高必须是2的倍数，如果报错，可以把device换为GPU
width = math.ceil(video_info['width']/2)*2
height = math.ceil(video_info['height']/2)*2
@@ -84,24 +94,45 @@ class StitchingVideo:
'vertical': f'{width}:-1',
}.get(stitching_type, f'{width}:-1')
print(f">>loop_count:{loop_count}")
if video1_audio or video2_audio:
#-map 1:a 指定使用第二个视频的音频流
command = f'ffmpeg {use_cuvid} -i {video1_path} -i {video2_path} -filter_complex "[1:v]scale={scale}[vid2];[0:v][vid2]{tack_type}=inputs=2[v]" -map "[v]" -map {use_audio}:a? {use_encoder} -c:a aac -strict experimental {output_path}'
command = f'ffmpeg "-y" {use_cuvid} -i "{video1_path}" -stream_loop {loop_count-1} -i "{video2_path}" -filter_complex "[0:v]setpts=PTS-STARTPTS[ref];[1:v]scale={scale},fps={fps},setpts=PTS-STARTPTS[vid2];[ref][vid2]{tack_type}=inputs=2[v]" -map "[v]" -map {use_audio}:a? {use_encoder} -c:a aac -strict experimental -t {duration} "{output_path}"'
else:
command = f'ffmpeg {use_cuvid} -i {video1_path} -i {video2_path} -filter_complex "[1:v]scale={scale}[vid2];[0:v][vid2]{tack_type}=inputs=2[v]" -map "[v]" {use_encoder} {output_path}'
command = f'ffmpeg "-y" {use_cuvid} -i "{video1_path}" -i "{video2_path}" -filter_complex "[0:v]setpts=PTS-STARTPTS[ref];[1:v]scale={scale},fps={fps}[vid2];[vid2]loop=loop={loop_count}:size=1:start=0,setpts=N/({fps})/TB[looped];[ref][looped]{tack_type}=inputs=2[v]" -map "[v]" {use_encoder} -t {duration} "{output_path}"'
print(command)
# 执行命令并检查错误
result = subprocess.run(command, stderr=subprocess.PIPE, stdout=subprocess.PIPE)
# 构建滤镜链
if scale_and_crop == "yes":
time.sleep(1)
if not os.path.isfile(output_path) and os.path.getsize(output_path)<=0:
print(f"output video：{output_path}不存在或文件内容为空!")
return (output_path,)
crop_video_path=os.path.join(os.path.dirname(output_path),"crop--"+final_output)
#ffmpeg "-y" -i Z:\sucai\video_right_test\20250520040751.mp4 -filter_complex "[0:v]split=2[bg][fg];[bg]scale=540:-1,setsar=1[scaled_bg];[scaled_bg]gblur=sigma=10[blurred];[blurred]scale=540:960:force_original_aspect_ratio=disable[bg_out];[fg]scale=540:-1,setsar=1[fg_out];[bg_out][fg_out]overlay=(W-w)/2:(H-h)/2[out];[out]scale=540:960:force_original_aspect_ratio=disable,setsar=1[final_out]" -map "[final_out]" -map 0:a -c:v h264_nvenc -c:a aac Z:\sucai\video_right_test\crop20250520040751.mp4
if stitching_type == "vertical":
command = f'ffmpeg "-y" -i "{output_path}" -filter_complex "[0:v]scale=-1:960[scaled];[scaled]crop=540:ih:(iw-540)/2:0[out]" -map "[out]" -map 0:a {use_encoder} -c:a aac "{crop_video_path}"'
else:
command = f'ffmpeg "-y" -i "{output_path}" -filter_complex "[0:v]split=2[bg][fg];[bg]scale=540:-1,setsar=1[scaled_bg];[scaled_bg]gblur=sigma=10[blurred];[blurred]scale=540:960:force_original_aspect_ratio=disable[bg_out];[fg]scale=540:-1,setsar=1[fg_out];[bg_out][fg_out]overlay=(W-w)/2:(H-h)/2[out];[out]scale=540:960:force_original_aspect_ratio=disable,setsar=1[final_out]" -map "[final_out]" -map 0:a {use_encoder} -c:a aac "{crop_video_path}"'
print(f">>FFmpeg Scale and crop::{command}")
result = subprocess.run(command, stderr=subprocess.PIPE, stdout=subprocess.PIPE)
output_path=crop_video_path
# 检查返回码
if result.returncode != 0:
# 如果有错误，输出错误信息
print(f"Error: {result.stderr.decode('utf-8')}")
raise ValueError(f"Error: {result.stderr.decode('utf-8')}")
print(f"Error: {result.stderr.decode('utf-8')}")
if device == "cuda":
print(f"***当前运算模式*[{device}]*************看下换成CPU重新执行，是否解决因编码问题的报错！********")
self.stitching_video(video1_path, video2_path,"cpu",use_audio,stitching_type,os.path.dirname(output_path),scale_and_crop)
else:
# 输出标准输出信息
print(result.stdout)
print(f">>FFmpeg 执行完毕！Completed!\t stdout: {result.stdout}")
return (output_path,)
except Exception as e:
#print("******[2]***********看下是否执行到这里！")
raise ValueError(e)