影视后期特效生成：在像素炼金术之外，重构“所见即所得”的工业逻辑

{
  "shot_042_explosion_v3": {
    "plugin": "Trapcode Particular",
    "params": {
      "particle_count": 5000,
      "turbulence_field": 0.75,
      "gravity": -9.8,
      "color_gradient": ["#FF4400", "#FFAA00", "#FFFFFF"],
      "size_over_life": [0.2, 1.5, 0.1]
    },
    "render_time_seconds": 187,
    "director_feedback": "爆炸扩散速度偏慢，需要更猛烈"
  }
}

import json
import datetime

def save_snapshot(shot_id, params_dict, feedback=""):
    snapshot = {
        "shot_id": shot_id,
        "timestamp": datetime.datetime.now().isoformat(),
        "params": params_dict,
        "feedback": feedback
    }
    filename = f"snapshots/{shot_id}_{datetime.datetime.now().strftime('%Y%m%d_%H%M')}.json"
    with open(filename, 'w') as f:
        json.dump(snapshot, f, indent=2)
    return filename

# 使用示例：每次手动调参后，调用此函数归档
# save_snapshot("shot_042", current_params, "导演反馈：加大粒子拖尾")

// AE表达式 —— 生成动态能量波纹遮罩
// 将此表达式挂在某个效果的"Opacity"或"Mask Feather"属性上

const centerX = thisComp.width / 2;
const centerY = thisComp.height / 2;
const dist = length(transform.position - [centerX, centerY]);
const wave = Math.sin(dist * 0.05 - time * 5) * 0.5 + 0.5;
// 让波纹从中心向外扩散，并随时间脉冲
const falloff = 1 - (dist / (thisComp.width * 0.8));
const result = Math.max(0, wave * falloff * 100);
result; // 返回0-100的值，控制透明度或强度

import os
import time
import subprocess
from glob import glob

RENDER_QUEUE_FOLDER = "./render_queue/"
RENDER_OUTPUT_FOLDER = "./rendered_outputs/"
LOG_FILE = "./render_log.txt"

def process_queue():
    while True:
        # 扫描队列文件夹中的.aep或.nk文件
        jobs = glob(os.path.join(RENDER_QUEUE_FOLDER, "*.aep"))
        if not jobs:
            print("队列空闲，等待5分钟再次扫描...")
            time.sleep(300)
            continue
        
        for job_file in jobs:
            print(f"开始渲染: {job_file}")
            # 调用AE的命令行渲染引擎（aerender）
            # -project 指定工程文件， -comp 指定合成名称， -output 指定输出路径
            output_name = os.path.basename(job_file).replace(".aep", "_output.mov")
            output_path = os.path.join(RENDER_OUTPUT_FOLDER, output_name)
            
            cmd = f'aerender -project "{job_file}" -comp "Main" -output "{output_path}"'
            result = subprocess.run(cmd, shell=True, capture_output=True, text=True)
            
            # 记录渲染日志
            with open(LOG_FILE, 'a') as log:
                log.write(f"{time.ctime()} | {job_file} | 耗时: {result} \n")
            
            # 渲染完成后移走源文件，防止重复渲染
            os.rename(job_file, job_file + ".done")
            
        time.sleep(10)  # 短暂休眠后继续扫描

if __name__ == "__main__":
    process_queue()

import cv2
import numpy as np
from collections import Counter

def analyze_dominant_colors(video_frame_path, top_k=3):
    """提取画面中最主要的k种颜色（RGB）"""
    img = cv2.imread(video_frame_path)
    img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
    # 降采样 + 聚类简化计算
    pixels = img.reshape(-1, 3)
    # 颜色量化：将相近色合并
    quantized = (pixels // 32) * 32  # 降低精度以聚类
    color_counts = Counter(map(tuple, quantized))
    dominant = color_counts.most_common(top_k)
    
    # 计算互补色 (255 - R, 255 - G, 255 - B)
    complements = [(255-r, 255-g, 255-b) for (r,g,b), count in dominant]
    return dominant, complements

# 实战场景：读取当前镜头第一帧，建议特效光效颜色
colors, comps = analyze_dominant_colors("shot_023_frame_001.png")
print(f"画面主色: {colors}")
print(f"建议特效光效色: {comps}")  # 这些颜色能自然融入画面并突出主体