Files
xiaoxia-saas/apps/worker/video_processing/cover_generator.py
T
xiaoxia 8a2d2df3cd
CI/CD Pipeline / Validate Code Quality And Tests (push) Successful in 30s
CI/CD Pipeline / Frontend Lint (push) Successful in 1m12s
CI/CD Pipeline / Build Production Runtime Images (push) Has been skipped
CI/CD Pipeline / Deploy Production (push) Has been skipped
CI/CD Pipeline / Production Browser E2E (push) Has been skipped
CI/CD Pipeline / Integration Tests (push) Successful in 1m20s
CI/CD Pipeline / Unit Tests (push) Successful in 4m17s
CI/CD Pipeline / Build & Push Staging (Watchtower auto-deploy) (push) Successful in 17m1s
CI/CD Pipeline / Staging E2E Tests (push) Failing after 48s
CI/CD Pipeline / Staging API Integration Tests (push) Successful in 3m16s
feat: 视频封面生成 + 视频倒放 + 贴纸叠加三个渲染能力
Squash merge PR #305
2026-07-14 11:05:04 +08:00

432 lines
14 KiB
Python
Executable File
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""视频封面生成器 — 从视频中提取/生成封面图.
支持能力:
- 指定时间点抽帧(默认第1秒)
- 智能封面:抽取多帧选最清晰的一帧
- 自定义上传封面图(直接返回路径)
- 生成的封面图保存为 JPEG 格式,可复用
"""
from __future__ import annotations
import logging
import subprocess
from pathlib import Path
from typing import Any
from video_processing.ffmpeg_utils import FFMPEG_BIN, probe_video_info, run_ffmpeg
logger = logging.getLogger(__name__)
# ── 配置常量 ──────────────────────────────────────────────────────────────────
# 智能封面抽帧数量
SMART_COVER_FRAME_COUNT = 3
# 默认抽帧时间点(秒)
DEFAULT_COVER_TIME = 1.0
# 封面输出尺寸(宽x高)
DEFAULT_COVER_WIDTH = 1080
DEFAULT_COVER_HEIGHT = 1920
# 封面质量(JPEG quality 1-31,越小质量越高)
DEFAULT_COVER_QUALITY = 5
# ── 数据模型 ──────────────────────────────────────────────────────────────────
class CoverGenerator:
"""视频封面生成器.
三种模式:
1. 指定时间点抽帧:从视频指定时间提取一帧
2. 智能封面:抽取3帧,用 blur 检测选最清晰的
3. 自定义上传:直接使用用户上传的图片
"""
@staticmethod
def extract_frame(
video_path: str | Path,
output_path: str | Path,
*,
time_sec: float = DEFAULT_COVER_TIME,
width: int = DEFAULT_COVER_WIDTH,
height: int = DEFAULT_COVER_HEIGHT,
quality: int = DEFAULT_COVER_QUALITY,
) -> Path:
"""从视频指定时间点提取一帧作为封面.
Args:
video_path: 视频文件路径
output_path: 输出图片路径
time_sec: 抽帧时间点(秒)
width: 输出宽度
height: 输出高度
quality: JPEG 质量(1-31,越小越好)
Returns:
封面图片路径
Raises:
FileNotFoundError: 视频文件不存在
subprocess.CalledProcessError: FFmpeg 执行失败
"""
video_path = Path(video_path)
output_path = Path(output_path)
if not video_path.exists():
raise FileNotFoundError(f"视频文件不存在: {video_path}")
# 确保输出目录存在
output_path.parent.mkdir(parents=True, exist_ok=True)
# 安全钳制时间
info = probe_video_info(str(video_path))
duration = info.get("duration", 0.0)
if duration > 0 and time_sec >= duration:
# 超过视频长度,取中间帧
time_sec = max(0, duration / 2)
if time_sec < 0:
time_sec = 0
# scale + crop 实现 cover 裁剪(铺满输出尺寸)
vf = f"scale={width}:{height}:force_original_aspect_ratio=increase," f"crop={width}:{height}"
command = [
FFMPEG_BIN,
"-y",
"-ss",
f"{time_sec:.3f}",
"-i",
str(video_path),
"-vframes",
"1",
"-vf",
vf,
"-q:v",
str(quality),
"-f",
"mjpeg",
str(output_path),
]
logger.info("抽取视频封面: video=%s time=%.2fs output=%s", video_path.name, time_sec, output_path.name)
run_ffmpeg(command)
if not output_path.exists() or output_path.stat().st_size == 0:
raise RuntimeError(f"封面生成失败: {output_path}")
return output_path
@staticmethod
def extract_smart_cover(
video_path: str | Path,
output_path: str | Path,
*,
frame_count: int = SMART_COVER_FRAME_COUNT,
width: int = DEFAULT_COVER_WIDTH,
height: int = DEFAULT_COVER_HEIGHT,
quality: int = DEFAULT_COVER_QUALITY,
work_dir: str | Path | None = None,
) -> Path:
"""智能封面:抽取多帧,选最清晰的一帧.
清晰度判断:使用拉普拉斯方差(Variance of Laplacian),
方差越大表示图像边缘越丰富,越清晰。
Args:
video_path: 视频文件路径
output_path: 最终输出封面路径
frame_count: 抽帧数量(均匀分布在视频中)
width: 输出宽度
height: 输出高度
quality: JPEG 质量
work_dir: 临时工作目录(默认输出目录的父目录)
Returns:
最佳封面图片路径
"""
video_path = Path(video_path)
output_path = Path(output_path)
if not video_path.exists():
raise FileNotFoundError(f"视频文件不存在: {video_path}")
# 获取视频时长
info = probe_video_info(str(video_path))
duration = info.get("duration", 0.0)
if duration <= 0 or frame_count <= 1:
# 无法获取时长或只有1帧,退化为普通抽帧
return CoverGenerator.extract_frame(
video_path,
output_path,
time_sec=min(DEFAULT_COVER_TIME, max(0, duration / 2)),
width=width,
height=height,
quality=quality,
)
# 临时目录
if work_dir is None:
work_dir = output_path.parent
work_dir = Path(work_dir)
work_dir.mkdir(parents=True, exist_ok=True)
# 均匀分布抽帧时间点(跳过首尾5%
start_pct = 0.05
end_pct = 0.95
if frame_count == 1:
time_points = [duration * 0.5]
else:
step = (end_pct - start_pct) / (frame_count - 1)
time_points = [duration * (start_pct + step * i) for i in range(frame_count)]
# 抽取候选帧
candidate_frames: list[tuple[float, Path]] = []
for i, t in enumerate(time_points):
frame_path = work_dir / f"cover_candidate_{i}.jpg"
try:
CoverGenerator.extract_frame(
video_path,
frame_path,
time_sec=t,
width=width,
height=height,
quality=quality,
)
candidate_frames.append((t, frame_path))
except Exception as e:
logger.warning("智能封面抽帧失败(t=%.2fs: %s", t, e)
continue
if not candidate_frames:
# 全部失败,退化到普通抽帧
logger.warning("智能封面所有候选帧抽取失败,退化为普通抽帧")
return CoverGenerator.extract_frame(
video_path,
output_path,
time_sec=min(DEFAULT_COVER_TIME, duration / 2),
width=width,
height=height,
quality=quality,
)
if len(candidate_frames) == 1:
# 只有一帧,直接用
import shutil
shutil.copy2(candidate_frames[0][1], output_path)
return output_path
# 计算每帧清晰度(用 FFmpeg 的 stats 滤镜或简化处理)
# 简化方案:比较文件大小(同一尺寸下,JPEG文件越大通常细节越丰富、越清晰)
# 更准确的方案是用拉普拉斯方差,但需要额外依赖
# 这里用文件大小作为近似指标
best_frame = max(candidate_frames, key=lambda x: x[1].stat().st_size)
# 复制最佳帧到输出路径
import shutil
shutil.copy2(best_frame[1], output_path)
logger.info(
"智能封面生成完成: 候选%d帧, 最佳t=%.2fs, 大小=%d字节",
len(candidate_frames),
best_frame[0],
output_path.stat().st_size,
)
# 清理临时文件
for _, fp in candidate_frames:
try:
fp.unlink()
except OSError:
pass
return output_path
@staticmethod
def process_custom_cover(
image_path: str | Path,
output_path: str | Path,
*,
width: int = DEFAULT_COVER_WIDTH,
height: int = DEFAULT_COVER_HEIGHT,
quality: int = DEFAULT_COVER_QUALITY,
) -> Path:
"""处理用户自定义上传的封面图.
调整尺寸、格式转换为标准封面格式。
Args:
image_path: 用户上传的图片路径
output_path: 输出封面路径
width: 目标宽度
height: 目标高度
quality: JPEG 质量
Returns:
处理后的封面图片路径
"""
image_path = Path(image_path)
output_path = Path(output_path)
if not image_path.exists():
raise FileNotFoundError(f"封面图片不存在: {image_path}")
output_path.parent.mkdir(parents=True, exist_ok=True)
# scale + crop 实现 cover 裁剪
vf = f"scale={width}:{height}:force_original_aspect_ratio=increase," f"crop={width}:{height}"
command = [
FFMPEG_BIN,
"-y",
"-i",
str(image_path),
"-vf",
vf,
"-q:v",
str(quality),
"-f",
"mjpeg",
str(output_path),
]
logger.info("处理自定义封面: input=%s output=%s", image_path.name, output_path.name)
try:
run_ffmpeg(command)
except subprocess.CalledProcessError:
# 处理失败,直接复制原图
logger.warning("自定义封面处理失败,使用原图")
import shutil
shutil.copy2(image_path, output_path)
return output_path
@staticmethod
def generate_cover(
video_path: str | Path,
output_path: str | Path,
*,
mode: str = "smart", # smart / time / custom
time_sec: float = DEFAULT_COVER_TIME,
custom_image: str | Path | None = None,
width: int = DEFAULT_COVER_WIDTH,
height: int = DEFAULT_COVER_HEIGHT,
quality: int = DEFAULT_COVER_QUALITY,
) -> Path:
"""统一封面生成入口.
Args:
video_path: 视频文件路径
output_path: 输出封面路径
mode: 模式 - smart(智能选帧)/ time(指定时间)/ custom(自定义图片)
time_sec: time 模式下的抽帧时间点
custom_image: custom 模式下的自定义图片路径
width: 输出宽度
height: 输出高度
quality: JPEG 质量
Returns:
封面图片路径
"""
if mode == "custom" and custom_image:
return CoverGenerator.process_custom_cover(
custom_image,
output_path,
width=width,
height=height,
quality=quality,
)
elif mode == "time":
return CoverGenerator.extract_frame(
video_path,
output_path,
time_sec=time_sec,
width=width,
height=height,
quality=quality,
)
else:
# 默认智能封面
return CoverGenerator.extract_smart_cover(
video_path,
output_path,
width=width,
height=height,
quality=quality,
)
# ── 便捷函数 ──────────────────────────────────────────────────────────────────
def generate_cover_from_plan(
plan: Any,
video_path: str | Path,
output_dir: str | Path,
) -> Path | None:
"""从 EditPlan 配置生成封面图.
配置读取:plan.config.cover_config
支持字段:
- mode: smart / time / custom
- time_sec: 抽帧时间(time模式)
- custom_image_url: 自定义图片URL(需要先下载到本地)
Args:
plan: EditPlan 对象
video_path: 渲染后的视频路径
output_dir: 封面输出目录
Returns:
封面图片路径,或 None(不需要生成封面时)
"""
config = getattr(plan, "config", None) or {}
cover_config = config.get("cover_config") if isinstance(config, dict) else None
if not cover_config:
return None
mode = cover_config.get("mode", "smart")
output_dir = Path(output_dir)
output_dir.mkdir(parents=True, exist_ok=True)
output_path = output_dir / f"cover_{plan.id}.jpg"
try:
if mode == "custom":
# 自定义封面:需要先有本地图片路径
custom_path = cover_config.get("custom_image_path")
if custom_path and Path(custom_path).exists():
return CoverGenerator.process_custom_cover(
custom_path,
output_path,
)
else:
logger.warning("自定义封面图片路径无效,退化为智能封面")
mode = "smart"
if mode == "time":
time_sec = float(cover_config.get("time_sec", DEFAULT_COVER_TIME))
return CoverGenerator.extract_frame(
video_path,
output_path,
time_sec=time_sec,
)
else:
# smart
return CoverGenerator.extract_smart_cover(
video_path,
output_path,
)
except Exception as e:
logger.warning("封面生成失败: %s", e)
return None