FFmpeg Cut Video at Keyframe Automatically

This article explains how to write an FFmpeg command that automatically cuts a video at the nearest keyframe. By using input seeking combined with stream copying, you can perform lightning-fast video cuts that snap directly to keyframes without the need for time-consuming re-encoding.

The Keyframe Cutting Command

To cut a video at the nearest keyframe without re-encoding, use the following FFmpeg command structure:

ffmpeg -ss 00:01:30 -i input.mp4 -t 00:00:45 -c copy output.mp4

How It Works

• -ss 00:01:30 (Input Seeking): Placing the start time parameter before the input file (-i) is crucial. This tells FFmpeg to use “input seeking.” FFmpeg will search the input file and automatically snap to the nearest keyframe (I-frame) just before or at the specified timestamp (1 minute and 30 seconds).
• -i input.mp4: Specifies your source video file.
• -t 00:00:45: Specifies the duration of the cut (45 seconds in this example). Because input seeking resets the video timeline to zero at the seek point, using -t is the most reliable way to define how long the cut should be.
• -c copy: This instructs FFmpeg to copy the video and audio streams directly without re-encoding them. This preservation of the original streams is what makes the process nearly instantaneous and lossless.
• output.mp4: The name of the exported video file.

Why Keyframes Matter

Video files are compressed using “keyframes” (which contain complete image data) and “interframes” (which only contain the differences between frames).

If you attempt to cut a video at a non-keyframe while copying the stream (-c copy), the player will not have a full image to render at the very beginning of the clip. This results in a frozen screen, black frames, or corrupted video artifacts until the player reaches the next keyframe.

By placing the -ss flag before -i, FFmpeg automatically seeks to the closest preceding keyframe, ensuring the output video starts with a clean, fully-rendered frame.

Keyframe Cutting vs. Accurate Cutting

Because this method snaps to the nearest existing keyframe, the cut may not happen at the exact millisecond you specified.

• For fast, lossless cuts: Use the command above. It takes seconds to process because it does not re-encode the video.
• For frame-accurate cuts: If you need to cut at an exact frame that is not a keyframe, you must re-encode the video. This takes longer but allows for pixel-perfect precision. To do this, replace -c copy with a video encoder, like so:

ffmpeg -i input.mp4 -ss 00:01:30 -t 00:00:45 -c:v libx264 -c:a aac output.mp4