调整大小与压缩
尺寸自动从源视频填充。清空字段或设为 -1 可保持宽高比。CRF 越低画质越高。
移除音频轨道
Output will be the same video with the audio stream stripped.
倒放
Plays video and audio backwards. The whole clip is buffered in memory - keep clips under ~2 minutes for best results.
淡入 / 淡出
设为 0 可跳过对应方向。在裁剪后应用。
自动字幕（Whisper）
大多数视频推荐使用 Base 模型。首次使用时下载并缓存。
转录文本（可编辑）
全程 100% 本地处理：音频提取后由 Transformers.js 调用本地 Whisper 转录，并以软字幕方式嵌入。数据零泄露。
音量
1.0 = 原始 · >1 = 增大 · <1 = 减小 · 0 = 静音。音频重新编码；视频流直接复制。
循环 / 重复
流直接复制，无需重新编码。片段连续播放 N 次。此操作忽略裁剪设置。
Logo / 图片叠加
带 Alpha 通道的 PNG 效果最佳。视频需重新编码以保留叠加效果。
混合音频（背景音乐）
音乐若短于视频则自动循环。视频流直接复制。
合并视频（连接片段）
Uses the concat filter - different resolutions and codecs are handled by re-encoding to H.264/AAC. Trim applies to the first clip only.
左右对比
Both clips are scaled to the same height (horizontal) or width (vertical) then stacked. Re-encodes to H.264/AAC. Trim is ignored.
画中画
Overlay video loops if shorter than the main clip. Audio from the main clip is preserved. Trim applies to the main clip.
媒体信息
加载视频文件以查看属性。
点击 处理视频 进行深度扫描 and log full codec & stream details below.
Raw FFmpeg
► Example Commands
--
Args are split on whitespace. Wrap values containing spaces in double quotes, e.g. -vf "scale=1280:-2,hflip"
Pad to Aspect Ratio
Video is scaled down to fit; empty space is filled with the chosen color. Re-encodes to H.264.
Normalize Audio (EBU R128)
Uses the loudnorm filter. Video stream is copied without re-encoding.
Denoise (hqdn3d)
Applies spatial + temporal noise reduction via hqdn3d. Higher strength = smoother but softer image.
回旋播放
Plays the clip forward then immediately in reverse — classic boomerang loop. Trim is respected. Audio is removed from the output.
锐化 / 模糊
Sharpen uses unsharp; blur uses boxblur. Audio is copied.
Operation Stack
No operations yet — configure an op above and click “Add to Stack”.
(stack is empty)
操作按列出的顺序作为一个滤镜链执行并进行一次重新编码。裁剪（如已设置）会首先应用。
Estimated output size
—
Output ready: