Frame-Accurate Video Cutting with FFmpeg

Cutting videos with frame-level precision in FFmpeg requires understanding the difference between fast seeking (input seeking) and accurate seeking (output seeking). This guide explains how to combine seeking parameters and re-encoding to split or cut videos at exact frames, avoiding frozen frames, black screens, or audio desync at the cut points.

Understanding the Keyframe Limitation

To cut a video with frame accuracy, you must understand how video compression works. Videos are made of keyframes (I-frames), which contain a complete image, and inter-frames (P-frames and B-frames), which only store the differences between frames.

If you cut a video using the stream copy mode (-c copy), FFmpeg can only cut at the nearest keyframe. If your specified cut time falls on a P-frame or B-frame, the video will start with a frozen or black screen until the next keyframe is reached.

To achieve true frame-accuracy, you must re-encode the video. Re-encoding allows FFmpeg to generate a new keyframe at the exact timestamp you specify.

This is the most reliable method for cutting a video at an exact frame. By placing the seek parameter (-ss) before the input, FFmpeg quickly seeks to the closest keyframe before your target time, and then decodes the remaining frames up to your start point. By re-encoding, it creates a clean new keyframe at the cut.

The Command:

ffmpeg -ss 00:01:30 -i input.mp4 -to 00:02:15 -c:v libx264 -c:a aac output.mp4

Parameter Breakdown:

  • -ss 00:01:30: The start time of the cut. Placing this before -i enables fast, frame-accurate seeking because FFmpeg will decode the video from the nearest prior keyframe.
  • -i input.mp4: The input video file.
  • -to 00:02:15: The end time of the cut. Alternatively, you can use -t 45 to specify a duration of 45 seconds.
  • -c:v libx264 -c:a aac: Re-encodes the video to H.264 and audio to AAC, which forces the creation of a new keyframe at the exact start point.

Method 2: Fast seeking with "Smart Cut" (Experimental)

If you have a very long video and do not want to re-encode the entire output, you can use a hybrid approach. This seeks quickly to the nearest keyframe, copies the majority of the video stream, but re-encodes only the frames at the cut point.

While FFmpeg does not have a native "smart cut" flag, you can achieve a similar result by splitting the process:

  1. Re-encode only the small segment from your cut point to the next keyframe.
  2. Copy the middle section without re-encoding.
  3. Concatenate the segments back together.

For most users, Method 1 is highly recommended because modern CPUs can re-encode short video clips quickly, ensuring 100% compatibility and eliminating audio/video sync issues.

Key Best Practices

  • Always put -ss before -i: When placed after the input, FFmpeg decodes every frame from the beginning of the video until it reaches the start time, which is incredibly slow for large files.
  • Use -to for timestamps, -t for duration: If you seek to 10 seconds (-ss 10) and want a 5-second clip, use -t 5. If you want to stop at the 15-second mark of the original video, use -to 15.