"""
Video Manager Advanced (合规用途) - Optimized Version
- 功能：合规的视频美化（中间主视频 + 两侧填充）、多层去重检测、并行处理、HTML 报告。
- 优化点：
- 增强去重精度：动态阈值调整、多级去重校验。
- 性能优化：并行帧提取、内存流式计算哈希。
- 音频指纹：添加降级方案（波形哈希）。
- 宽幅合成：支持动态模糊、多层叠加（侧边内容）。
- 报告：动态多帧缩略图、交互按钮（模拟删除）。
- 清理：自动删除临时目录、错误日志记录。
- 严禁用于规避平台检测、传播未授权或侵权内容。
"""
import os
import sys
import shutil
import subprocess
import time
import csv
import tempfile
import math
from
pathlib import
Path
from concurrent.futures import ThreadPoolExecutor, as_completed
from multiprocessing import cpu_count
from PIL import Image
import imagehash
import hashlib
import base64
from tqdm import tqdm
import PySimpleGUI as sg
# ----------------------------
# 配置（可按需修改）
# ----------------------------
FFMPEG = "ffmpeg" # or full
path to ffmpeg
FFPROBE = "ffprobe"
FPCALC = "fpcalc" # optional
SUPPORTED_EXT = [".mp4", ".mov", ".mkv", ".ts", ".webm"]
TMP_PREFIX = "vma_"
ERROR_LOG = "vma_
error.log"
# ----------------------------
# 工具函数
# ----------------------------
def run_quiet(cmd)
:
"""Run subprocess and swallow output (returns CompletedProcess)"""
return subprocess.run(cmd, stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
def log_
error(e, context="")
:
"""记录错误到日志文件"""
with open(
ERROR_LOG, "a", encoding="utf-8") as f
:
f.write(f"{time.ctime()} [{context}]
: {str(e)}\n")
def file_sha256(
path)
:
h = hashlib.sha256()
with open(
path, "rb") as f
:
for chunk in iter(lambda
: f.read(1 << 20), b"")
:
h.update(chunk)
return h.hexdigest()
def ffprobe_duration(
path)
:
try
:
out = subprocess.check_output([FFPROBE, "-v", "
error", "-show_entries", "format=duration",
"-
of", "default=noprint_wrappers=1
:nokey=1", str(
path)],
stderr=subprocess.DEVNULL).decode().strip()
return float(out)
except
Exception as e
:
log_
error(e, "ffprobe_duration")
return None
def extract_frames(video_
path, out_dir, fps=1, max_frames=None)
:
"""
Extract frames at given fps into out_dir as frame_0001.jpg ...
If fps
is float, it's frames per second. Use -vf fps=...
"""
os.makedirs(out_dir, ex
ist_ok=True)
if fps <= 0
:
return []
vf = f"fps={fps}"
cmd = [FFMPEG, "-y", "-i", str(video_
path), "-vf", vf, "-q
:v", "2", os.
path.join(out_dir, "frame_%05d.jpg")]
proc = run_quiet(cmd)
if proc.returncode != 0
:
log_
error(proc.stderr, "extract_frames")
return []
files = sorted(
Path(out_dir).glob("frame_*.jpg"))
if max_frames and len(files) > max_frames
:
files = files[
:max_frames]
return [str(p) for p in files]
def extract_mid_frame(video_
path, out_image)
:
dur = ffprobe_duration(video_
path)
if dur
is None
:
ss = 1.0
else
:
ss = max(0.5, dur / 2.0)
cmd = [FFMPEG, "-y", "-ss", str(ss), "-i", str(video_
path), "-frames
:v", "1", "-q
:v", "2", str(out_image)]
proc = run_quiet(cmd)
return os.
path.ex
ists(out_image)
def compute_phash_from_image(img_
path)
:
try
:
with Image.open(img_
path).convert("RGB") as img
:
return str(imagehash.phash(img))
except
Exception as e
:
log_
error(e, "compute_phash_from_image")
return None
def compute_phash_streaming(frame_
paths)
:
"""流式计算哈希，避免内存过高"""
phashes = []
for f in frame_
paths
:
ph = compute_phash_from_image(f)
if ph
:
phashes.append(ph)
return phashes
def hamming_d
istance(ph1, ph2)
:
try
:
if
isinstance(ph1, str)
: ph1 = imagehash.hex_to_hash(ph1)
if
isinstance(ph2, str)
: ph2 = imagehash.hex_to_hash(ph2)
return ph1 - ph2
except
Exception:
return 9999
def audio_fingerprint(video_
path)
:
"""音频指纹：优先 fpcalc，降级到波形哈希"""
if shutil.which(FPCALC)
:
try
:
tmpwav = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
tmpwav.close()
cmd1 = [FFMPEG, "-y", "-i", str(video_
path), "-vn", "-acodec", "pcm_s16le", "-ar", "44100", "-ac", "2", tmpwav.name]
subprocess.run(cmd1, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
res = subprocess.run([FPCALC, "-raw", tmpwav.name], capture_output=True, text=True)
os.unlink(tmpwav.name)
if res.returncode == 0
:
return res.stdout.strip()
except
Exception as e
:
log_
error(e, "audio_fingerprint_fpcalc")
# 降级：简单波形哈希
try
:
cmd = [FFMPEG, "-i", str(video_
path), "-vn", "-af", "volumedetect", "-f", "null", "-"]
out = subprocess.run(cmd, capture_output=True, text=True).stderr
if "mean_volume" in out
:
return hashlib.md5(out.encode()).hexdigest()
except
Exception as e
:
log_
error(e, "audio_fingerprint_fallback")
return None
def
dynamic_blur(video_
path, frames_per_sec=0.5, max_frames=2)
:
"""动态计算模糊值，根据帧差异"""
tmpdir = tempfile.mkdtemp(prefix=TMP_PREFIX)
frames = extract_frames(video_
path, tmpdir, fps=frames_per_sec, max_frames=max_frames)
if len(frames) < 2
:
shutil.rmtree(tmpdir, ignore_
errors=True)
return 18 # 默认
phashes = compute_phash_streaming(frames)
diffs = [hamming_d
istance(phashes[i], phashes[i+1]) for i in range(len(phashes)-1)]
avg_diff = sum(diffs) / len(diffs) if diffs else 0
blur = min(30, max(10, int(avg_diff * 2)))
shutil.rmtree(tmpdir, ignore_
errors=True)
return blur
def generate_wide_version(src, dest, out_w=1920, out_h=1080, blur=None, side_content=None)
:
"""宽幅合成：支持动态模糊、多层叠加"""
if blur
is None
:
blur =
dynamic_blur(src)
fc = f"[0
:v]scale={out_w}
:{out_h},boxblur={blur}
:1[bg];[0
:v]scale=-1
:{out_h}[fg]"
if side_content
:
# 假设 side_content 是额外输入文件路径
fc = f"[0
:v]scale={out_w}
:{out_h},boxblur={blur}
:1[bg];" \
f"[1
:v]scale={out_w/4}
:{out_h}[side];[bg][side]overlay=x=0
:y=0[bg2];" \
f"[0
:v]scale=-1
:{out_h}[fg];[bg2][fg]overlay=(W-w)/2
:(H-h)/2"
cmd = [FFMPEG, "-y", "-i", str(src), "-i", str(side_content), "-filter_complex", fc, "-c
:a", "copy", str(dest)]
else
:
cmd = [FFMPEG, "-y", "-i", str(src), "-filter_complex", fc, "-c
:a", "copy", str(dest)]
proc = run_quiet(cmd)
return proc.returncode == 0
# ----------------------------
# 核心去重逻辑（并行）
# ----------------------------
def analyze_file(video_
path, tempdir, frames_per_sec=1, max_frames=5, use_audio=False)
:
rec = {"
path"
: str(video_
path), "sha256"
: None, "frames"
: [], "phashes"
: [], "audio_fp"
: None}
try
:
rec["sha256"] = file_sha256(video_
path)
fdir = os.
path.join(tempdir,
Path(video_
path).stem)
os.makedirs(fdir, ex
ist_ok=True)
frames = extract_frames(video_
path, fdir, fps=frames_per_sec, max_frames=max_frames)
rec["frames"] = frames
rec["phashes"] = compute_phash_streaming(frames)
if use_audio
:
rec["audio_fp"] = audio_fingerprint(video_
path)
except
Exception as e
:
log_
error(e, f"analyze_file
: {video_
path}")
return rec
def build_index(folder, workers=None, frames_per_sec=1, max_frames=5, use_audio=False, progress_callback=None)
:
folder =
Path(folder)
files = [p for p in folder.iterdir() if p.
is_file() and p.suffix.lower() in
SUPPORTED_EXT]
total = len(files)
tmpdir = tempfile.mkdtemp(prefix=TMP_PREFIX)
results = []
workers = workers or max(1, cpu_count()//2)
with ThreadPoolExecutor(max_workers=workers) as ex
:
futures = {ex.submit(analyze_file, p, tmpdir, frames_per_sec, max_frames, use_audio)
: p for p in files}
for i, fut in enumerate(as_completed(futures))
:
try
:
r = fut.result()
results.append(r)
except
Exception as e
:
log_
error(e, "build_index")
if progress_callback
:
progress_callback(i+1, total)
return results, tmpdir
def
dynamic_threshold(phashes)
:
"""动态阈值：基于帧间差异"""
if len(phashes) < 2
:
return 8
diffs = [hamming_d
istance(phashes[i], phashes[i+1]) for i in range(len(phashes)-1)]
avg_diff = sum(diffs) / len(diffs) if diffs else 0
return max(8, int(avg_diff * 1.5))
def multi_level_dedup(rec_a, rec_b, phash_threshold=8,
require_frame_ratio=0.6)
:
"""多级去重：SHA256 + 帧哈希 + 音频"""
if rec_a["sha256"] == rec_b["sha256"]
:
return 1.0, True # exact match
audio_match = rec_a.get("audio_fp") == rec_b.get("audio_fp") if rec_a.get("audio_fp") and rec_b.get("audio_fp") else False
aps = rec_a.get("phashes", [])
bps = rec_b.get("phashes", [])
if
not aps or
not bps
:
return 0.0, audio_match
thresh_a =
dynamic_threshold(aps)
thresh_b =
dynamic_threshold(bps)
thresh = max(thresh_a, thresh_b, phash_threshold)
match_count = sum(1 for ph_a in aps for ph_b in bps if hamming_d
istance(ph_a, ph_b) <= thresh)
total_frames = max(1, min(len(aps), len(bps)))
ratio = match_count / (total_frames * max(len(aps), len(bps)) / total_frames) # normalized
return ratio, audio_match or (ratio >=
require_frame_ratio)
def compare_records(records, phash_threshold=8,
require_frame_ratio=0.6)
:
out = []
n = len(records)
for i in range(n)
:
for j in range(i+1, n)
:
a = records[i]; b = records[j]
ratio, audio_match = multi_level_dedup(a, b, phash_threshold,
require_frame_ratio)
if ratio >=
require_frame_ratio or audio_match
:
best_d
ist = min(hamming_d
istance(pa, pb) for pa in a["phashes"] for pb in b["phashes"]) if a["phashes"] and b["phashes"] else 999
out.append((a["
path"], b["
path"], best_d
ist, ratio, audio_match))
return out
# ----------------------------
# 报告生成（CSV + HTML preview）
# ----------------------------
def generate_csv_report(records, pairs, out_csv)
:
with open(out_csv, "w", newline="", encoding="utf-8") as f
:
w = csv.writer(f)
w.writerow(["
path", "sha256", "num_frames", "phashes", "audio_fp_present"])
for r in records
:
w.writerow([r["
path"], r["sha256"], len(r.get("frames", [])), ";".join(r.get("phashes", [])), bool(r.get("audio_fp"))])
w.writerow([])
w.writerow(["potential_duplicate_pairs"])
w.writerow(["
path_a", "
path_b", "best_phash_d
istance", "matching_ratio", "audio_match"])
for a,b,d,ratio,audio in pairs
:
w.writerow([a,b,d,ratio,audio])
def encode_multi_thumbnail(frame_
paths, maxw=360)
:
"""动态多帧缩略图：拼接前3帧"""
try
:
imgs = []
for f in frame_
paths[
:3]
:
with Image.open(f).convert("RGB") as img
:
imgs.append(img.resize((120, 120)))
if
not imgs
:
return ""
result = Image.new("RGB", (120 * len(imgs), 120))
for i, img in enumerate(imgs)
:
result.paste(img, (120 * i, 0))
from io import BytesIO
buf = BytesIO()
result.save(buf, format="JPEG", quality=75)
return base64.b64encode(buf.getvalue()).decode("ascii")
except
Exception as e
:
log_
error(e, "encode_multi_thumbnail")
return ""
def generate_html_report(records, pairs, out_html, tmpdir)
:
html = ["<html><head><meta charset='utf-8'><title>Dedup Report</title></head><body>"]
html.append("<h1>Video Manager Advanced - Report</h1>")
html.append("<h2>Files</h2><div style='d
isplay
:flex;flex-wrap
:wrap;'>")
for r in records
:
b64 = encode_multi_thumbnail(r["frames"]) if r["frames"] else ""
html.append("<div style='width
:380px;border
:1px solid #ddd;margin
:6px;padding
:6px;'>")
if b64
:
html.append(f"<img src='data
:image/jpeg;base64,{b64}' style='max-width
:360px;d
isplay
:block;margin-bottom
:4px;'/>")
html.append(f"<div style='font-size
:12px;word-break
:break-all;'>
Path: {r['
path']}</div>")
html.append(f"<div style='font-size
:12px;'>SHA256
: {r['sha256'][
:16]}...</div>")
html.append("</div>")
html.append("</div>")
html.append("<h2>Potential Duplicates</h2>")
html.append("<table border='1' cellpadding='6'><tr><th>File A</th><th>File B</th><th>best_d
ist</th><th>ratio</th><th>audio_match</th><th>Action</th></tr>")
for a,b,d,ratio,audio in pairs
:
html.append(f"<tr><td style='max-width
:300px'>{a}</td><td style='max-width
:300px'>{b}</td><td>{d}</td><td>{ratio
:.2f}</td><td>{audio}</td>"
f"<td><button onclick=\"if(confirm('Delete {a}?')) alert('Deleted (simulated)');\">Del A</button> "
f"<button onclick=\"if(confirm('Delete {b}?')) alert('Deleted (simulated)');\">Del B</button></td></tr>")
html.append("</table>")
html.append("<p>
Note
: Th
is tool does
NOT delete files automatically. Please review results manually.</p>")
html.append("</body></html>")
with open(out_html, "w", encoding="utf-8") as f
:
f.write("\n".join(html))
# ----------------------------
# GUI 主窗口（带进度/日志）
# ----------------------------
def run_gui()
:
sg.theme("SystemDefault")
layout = [
[sg.Text("Video Manager Advanced（合规用途）", font=("Helvetica", 14))],
[sg.Text("1) 扫描与去重")],
[sg.Text("Folder
:"), sg.Input(key="-FOLDER-"), sg.FolderBrowse()],
[sg.Text("Frames/sec (采样帧率)
:"), sg.Input("1", size=(6,1), key="-FPS-"),
sg.Text("Max frames per file
:"), sg.Input("5", size=(6,1), key="-MAXF-"),
sg.Text("phash阈值
:"), sg.Input("8", size=(6,1), key="-PH-")],
[sg.Checkbox("启用音频指纹(fpcalc)", key="-AF-", default=False),
sg.Text("并发线程
:"), sg.Input(str(max(1, cpu_count()//2)), size=(6,1), key="-WORK-")],
[sg.Button("开始扫描", key="-SCAN-"), sg.Button("停止", key="-STOP-"), sg.Text("", key="-STATUS-", size=(40,1))],
[sg.ProgressBar(100, orientation='h', size=(50, 20), key='-PROG-')],
[sg.HorizontalSeparator()],
[sg.Text("2) 合成宽屏视频（中间主视频 + 两侧模糊）")],
[sg.Text("Input Video
:"), sg.Input(key="-INVID-"), sg.FileBrowse(file_types=(("Video","*.mp4;*.mov;*.mkv"),))],
[sg.Text("Side Content (optional)
:"), sg.Input(key="-SIDE-",), sg.FileBrowse(file_types=(("Video/Image","*.mp4;*.jpg;*.png"),))],
[sg.Text("Output
:"), sg.Input("out_wide.mp4", key="-OUTVID-"), sg.FileSaveAs(file_types=(("MP4","*.mp4"),))],
[sg.Text("Out W
:"), sg.Input("1920", key="-W-", size=(6,1)), sg.Text("Out H
:"), sg.Input("1080", key="-H-", size=(6,1)),
sg.Text("Blur (0=auto)
:"), sg.Input("0", key="-BLUR-", size=(6,1))],
[sg.Button("合成视频", key="-MAKE-")],
[sg.HorizontalSeparator()],
[sg.Text("日志
:")],
[sg.Multiline(size=(100,10), key="-LOG-", autoscroll=True)],
[sg.Button("退出")]
]
window = sg.Window("VideoManagerAdv", layout, resizable=True, finalize=True)
scanning = False
tmpdir_global = None
while True
:
event, values = window.read(timeout=100)
if event in (sg.WIN_CLOSED, "退出")
:
if tmpdir_global and os.
path.ex
ists(tmpdir_global)
:
shutil.rmtree(tmpdir_global, ignore_
errors=True)
break
if event == "-MAKE-"
:
invid = values["-INVID-"]
outvid = values["-OUTVID-"]
side = values["-SIDE-"] or None
if
not invid or
not outvid
:
sg.popup("请指定输入与输出视频文件")
continue
try
:
ow = int(values["-W-"]); oh = int(values["-H-"]); blur = int(values["-BLUR-"])
if blur == 0
:
blur = None # auto
except
Exception:
sg.popup("宽高或模糊参数错误")
continue
window['-LOG-'].print(f"开始合成 {invid} -> {outvid} ...")
ok = generate_wide_version(invid, outvid, out_w=ow, out_h=oh, blur=blur, side_content=side)
window['-LOG-'].print("合成完成" if ok else "合成失败，请检查 ffmpeg")
if event == "-SCAN-"
:
folder = values["-FOLDER-"]
if
not folder or
not os.
path.
isdir(folder)
:
sg.popup("请选择有效文件夹")
continue
try
:
fps = float(values["-FPS-"])
maxf = int(values["-MAXF-"])
ph = int(values["-PH-"])
workers = int(values["-WORK-"])
use_audio = bool(values["-AF-"])
except
Exception:
sg.popup("参数错误")
continue
window['-STATUS-'].update("Scanning...")
window['-LOG-'].print(f"开始索引目录
: {folder}")
scanning = True
window['-PROG-'].update(0)
try
:
records, tmpdir = build_index(folder, workers=workers, frames_per_sec=fps, max_frames=maxf, use_audio=use_audio,
progress_callback=lambda done, total
: window['-PROG-'].update(int(done/total*100)))
tmpdir_global = tmpdir
window['-LOG-'].print(f"索引完成, 临时目录
: {tmpdir}")
window['-LOG-'].print("开始相似度比对 ...")
pairs = compare_records(records, phash_threshold=ph)
out_csv = os.
path.join(folder, "vma_report.csv")
out_html = os.
path.join(folder, "vma_report.html")
generate_csv_report(records, pairs, out_csv)
generate_html_report(records, pairs, out_html, tmpdir)
window['-LOG-'].print(f"报告生成
: {out_csv}, {out_html}")
window['-STATUS-'].update("完成")
window['-PROG-'].update(100)
except
Exception as e
:
log_
error(e, "-SCAN-")
window['-LOG-'].print("错误
: " + str(e))
window['-STATUS-'].update("错误")
scanning = False
if event == "-STOP-"
:
sg.popup("当前操作为短任务；若需中止正在运行的 ffmpeg，请在任务管理器中结束 ffmpeg 进程。")
window.close()
# ----------------------------
# CLI 支持（无 GUI 批量）
# ----------------------------
def run_cli(args)
:
import argparse
parser = argparse.ArgumentParser(description="Video Manager Advanced CLI")
parser.add_argument("folder", help="Folder to scan")
parser.add_argument("--fps", type=float, default=1.0, help="frames per second to sample")
parser.add_argument("--max-frames", type=int, default=5, help="max frames per file")
parser.add_argument("--workers", type=int, default=max(1, cpu_count()//2))
parser.add_argument("--phash-threshold", type=int, default=8)
parser.add_argument("--use-audio-fp", action="store_true")
parser.add_argument("--out-csv", default="vma_report.csv")
parser.add_argument("--out-html", default="vma_report.html")
ns = parser.parse_args(args)
print("Building index...")
records, tmp = build_index(ns.folder, workers=ns.workers, frames_per_sec=ns.fps, max_frames=ns.max_frames, use_audio=ns.use_audio_fp,
progress_callback=lambda done, total
: print(f"Progress
: {done}/{total}", end="\r"))
print("\nComparing records...")
pairs = compare_records(records, phash_threshold=ns.phash_threshold)
generate_csv_report(records, pairs, ns.out_csv)
generate_html_report(records, pairs, ns.out_html, tmp)
print("Reports
:", ns.out_csv, ns.out_html)
print("Temporary frames in
:", tmp)
print("
Note
: Tool does
NOT delete files automatically. Please review reports manually.")
# CLI 模式下清理临时目录
if tmp and os.
path.ex
ists(tmp)
:
shutil.rmtree(tmp, ignore_
errors=True)
# ----------------------------
# 入口
# ----------------------------
if __name__ == "__main__"
:
if len(sys.argv) > 1
:
# CLI mode
run_cli(sys.argv[1
:])
else
:
run_gui()
文章讲述了在使用Vite5.0.0构建Vue3项目时，由于不支持动态require(path)，导致vite.config.js报错。解决方案是将require替换为import。
扫一扫
31万+
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
