ADD 增加视频合并节点

PERF 生成视频支持输出尾帧用于生成下一段视频

__init__.py

@@ -9,7 +9,8 @@ from .nodes.text_nodes import StringEmptyJudgement, LoadText, RandomLineSelector
 from .nodes.util_nodes import LogToDB, TaskIdGenerate, TraverseFolder, UnloadAllModels, VodToLocalNode, \
     PlugAndPlayWebhook
 from .nodes.video_lipsync_nodes import HeyGemF2F, HeyGemF2FFromFile
-from .nodes.video_nodes import VideoCut, VideoCutByFramePoint, VideoChangeFPS, VideoStartPointDurationCompute
+from .nodes.video_nodes import VideoCut, VideoCutByFramePoint, VideoChangeFPS, VideoStartPointDurationCompute, \
+    VideoMerge
 
 NODE_CLASS_MAPPINGS = {
     "FaceOccDetect": FaceDetect,
@@ -47,7 +48,8 @@ NODE_CLASS_MAPPINGS = {
     "ModalEditCustom": ModalEditCustom,
     "ModalMidJourneyGenerateImage": ModalMidJourneyGenerateImage,
     "ModalMidJourneyDescribeImage": ModalMidJourneyDescribeImage,
-    "JMCustom": JMCustom
+    "JMCustom": JMCustom,
+    "VideoMerge": VideoMerge
 }
 
 NODE_DISPLAY_NAME_MAPPINGS = {
@@ -86,5 +88,6 @@ NODE_DISPLAY_NAME_MAPPINGS = {
     "ModalEditCustom": "自定义Prompt修改图片",
     "ModalMidJourneyGenerateImage": "Prompt修图",
     "ModalMidJourneyDescribeImage": "反推生图提示词",
-    "JMCustom": "Prompt生视频"
+    "JMCustom": "Prompt生视频",
+    "VideoMerge":"顺序合并视频"
 }

nodes/image_gesture_nodes.py

@@ -7,6 +7,7 @@ import time
 import uuid
 from time import sleep
 
+import cv2
 import folder_paths
 import numpy as np
 import requests
@@ -15,6 +16,7 @@ import yaml
 from PIL import Image
 from loguru import logger
 from qcloud_cos import CosConfig, CosS3Client
+from torchvision.transforms import transforms
 from tqdm import tqdm
 
 
@@ -119,12 +121,63 @@ class JMUtils:
 
     def tensor_to_io(srlf, tensor: torch.Tensor):
         # 转换为PIL图像
-        img = Image.fromarray(np.clip(255. * tensor.cpu().numpy().squeeze(), 0, 255).astype(np.uint8))
+        img = Image.fromarray(np.clip(255. * tensor.cpu().squeeze().numpy(), 0, 255).astype(np.uint8))
         image_data = io.BytesIO()
         img.save(image_data, format='PNG')
         image_data.seek(0)
         return image_data
 
+    def read_video_last_frame_to_tensor(self, video_path: str) -> torch.Tensor:
+        """
+        读取视频文件的最后一帧并将其转换为BCHW格式的PyTorch张量。
+
+        参数:
+            video_path (str): 视频文件的路径。
+
+        返回:
+            torch.Tensor: 形状为[1, H, W, C]的张量，其中H和W分别是视频帧的高度和宽度，通道顺序为RGB。
+
+        异常:
+            FileNotFoundError: 如果指定的视频文件不存在。
+            ValueError: 如果视频文件为空或无法读取帧。
+        """
+        # 打开视频文件
+        cap = cv2.VideoCapture(video_path)
+
+        if not cap.isOpened():
+            raise FileNotFoundError(f"无法打开视频文件: {video_path}")
+
+        # 获取视频总帧数
+        total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+
+        if total_frames == 0:
+            cap.release()
+            raise ValueError("视频文件为空或无法确定帧数")
+
+        # 设置读取位置到最后一帧
+        cap.set(cv2.CAP_PROP_POS_FRAMES, total_frames - 1)
+
+        # 读取最后一帧
+        ret, frame = cap.read()
+
+        # 释放资源
+        cap.release()
+
+        if not ret or frame is None:
+            raise ValueError(f"无法读取视频的最后一帧，可能视频已损坏")
+
+        # 转换BGR到RGB (OpenCV默认读取为BGR)
+        frame_rgb = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)
+
+        # 转换为PyTorch张量并调整维度为BCHW
+        transform = transforms.Compose([
+            transforms.ToTensor()  # 转换为[C, H, W]格式的张量，值范围从0到1
+        ])
+
+        tensor = transform(frame_rgb).unsqueeze(0).permute(0, 2, 3, 1)  # 添加批次维度，变为[1, H, W, C]
+
+        return tensor
+
     def download_video(self, url, timeout=30, retries=3, path=None):
         """下载视频到临时文件并返回文件路径"""
         for attempt in range(retries):
@@ -159,7 +212,7 @@ class JMUtils:
                         bar.update(size)
 
                 print(f"视频下载完成: {temp_path}")
-                return temp_path
+                return temp_path, self.read_video_last_frame_to_tensor(temp_path)
 
             except Exception as e:
                 print(f"下载错误 (尝试 {attempt + 1}/{retries}): {str(e)}")
@@ -336,8 +389,8 @@ class JMCustom:
             }
         }
 
-    RETURN_TYPES = ("STRING",)
+    RETURN_TYPES = ("STRING", "IMAGE",)
-    RETURN_NAMES = ("视频存储路径",)
+    RETURN_NAMES = ("视频存储路径", "视频最后一帧")
     FUNCTION = "gen"
     CATEGORY = "不忘科技-自定义节点🚩/视频/即梦"
 
@@ -368,5 +421,6 @@ class JMCustom:
             sleep(interval)
         if not job_data:
             raise Exception("即梦任务等待超时")
-        return (
+        video_path, last_scene = client.download_video(job_data, path=os.path.join(folder_paths.get_output_directory(),
-        client.download_video(job_data, path=os.path.join(folder_paths.get_output_directory(), f"{uuid.uuid4()}.mp4")),)
+                                                                                   f"{uuid.uuid4()}.mp4"))
+        return (video_path, last_scene,)

nodes/video_nodes.py

@@ -1,4 +1,5 @@
 import glob
+import json
 import os
 import re
 import shutil
@@ -6,8 +7,11 @@ import subprocess
 import traceback
 import uuid
 from datetime import datetime
+from pathlib import Path
+from typing import List
 
 import ffmpy
+import folder_paths
 import loguru
 import torchvision.io
 
@@ -22,7 +26,7 @@ class VideoCut:
         return {
             "required": {
                 "video_path": (
-                "STRING", {"placeholder": "X://insert/path/here.mp4", "vhs_path_extensions": video_extensions}),
+                    "STRING", {"placeholder": "X://insert/path/here.mp4", "vhs_path_extensions": video_extensions}),
                 "start": ("STRING", {"default": "00:00:00.000"}),
                 "end": ("STRING", {"default": "00:00:10.000"}),
             },
@@ -119,7 +123,8 @@ class VideoCut:
                 os.remove(files[0])
             except:
                 pass
-            return (video/255.0, {"waveform": audio, "sample_rate": info["audio_fps"]} if "audio_fps" in info else None,)
+            return (
+            video / 255.0, {"waveform": audio, "sample_rate": info["audio_fps"]} if "audio_fps" in info else None,)
         except:
             traceback.print_exc()
             raise Exception("Cut Failed")
@@ -133,7 +138,7 @@ class VideoCutByFramePoint:
         return {
             "required": {
                 "video_path": (
-                "STRING", {"placeholder": "X://insert/path/here.mp4", "vhs_path_extensions": video_extensions}),
+                    "STRING", {"placeholder": "X://insert/path/here.mp4", "vhs_path_extensions": video_extensions}),
                 "start_point": ("FLOAT", {"default": "0.0"}),
                 "duration": ("FLOAT", {"default": "10.0"}),
                 "fps": ("INT", {"default": "25"}),
@@ -234,7 +239,7 @@ class VideoCutByFramePoint:
                 os.remove(output)
             except:
                 pass
-            return (video/255.0, {"waveform": audio, "sample_rate": info["audio_fps"]},)
+            return (video / 255.0, {"waveform": audio, "sample_rate": info["audio_fps"]},)
         except:
             traceback.print_exc()
             raise Exception("Cut Failed")
@@ -248,7 +253,7 @@ class VideoChangeFPS:
         return {
             "required": {
                 "video_path": (
-                "STRING", {"placeholder": "X://insert/path/here.mp4", "vhs_path_extensions": video_extensions}),
+                    "STRING", {"placeholder": "X://insert/path/here.mp4", "vhs_path_extensions": video_extensions}),
                 "fps": ("INT", {"default": 30}),
             },
         }
@@ -392,3 +397,76 @@ class VideoStartPointDurationCompute:
         duration = duration + end_padding
         loguru.logger.info("audio duration with padding %.3f s" % duration)
         return (start_point, duration * fps,)
+
+
+def merge_videos(input_paths: List[str], output_path: str) -> str:
+    """
+    按顺序拼接多个视频文件到一个输出文件
+
+    参数:
+        input_paths: 视频文件路径列表，按拼接顺序排列
+        output_path: 输出视频文件路径
+    """
+    # 检查所有输入文件是否存在
+    for path in input_paths:
+        if not Path(path).exists():
+            raise FileNotFoundError(f"输入文件不存在: {path}")
+
+    # 创建临时文件列表
+    temp_filelist = os.path.join(os.path.dirname(__file__),"filelist.txt")
+    with open(temp_filelist, "w", encoding="utf-8") as f:
+        for path in input_paths:
+            # 处理路径中的引号和特殊字符
+            escaped_path = path.replace("'", r"'\''")
+            f.write(f"file '{escaped_path}'\n")
+
+    try:
+        # 使用ffmpeg执行拼接操作
+        cmd = [
+            "ffmpeg",
+            "-f", "concat",
+            "-safe", "0",
+            "-i", str(temp_filelist),
+            "-c", "copy",
+            output_path
+        ]
+
+        result = subprocess.run(
+            cmd,
+            capture_output=True,
+            text=True,
+            check=True
+        )
+
+        print(f"视频拼接成功，输出文件: {output_path}")
+        print("ffmpeg 输出:", result.stderr)
+        return output_path
+
+    except subprocess.CalledProcessError as e:
+        print(f"视频拼接失败: {e.stderr}")
+        raise e
+    finally:
+        if os.path.exists(temp_filelist):
+            os.remove(temp_filelist)
+
+
+class VideoMerge:
+    @classmethod
+    def INPUT_TYPES(cls):
+        return {
+            "required": {
+                "video_list": ("STRING", {"default": "[]"})
+            },
+        }
+
+    RETURN_TYPES = ("STRING",)
+    RETURN_NAMES = ("视频路径",)
+
+    FUNCTION = "process"
+
+    CATEGORY = "不忘科技-自定义节点🚩/视频"
+
+    def process(self, video_list):
+        if isinstance(video_list, str):
+            video_list = json.loads(video_list)
+        return (merge_videos(video_list, os.path.join(folder_paths.get_output_directory(), f"merged_{uuid.uuid4()}.mp4")),)