Removing Every Nth Frame in Raw Video Using FFmpeg

To remove every Nth frame from a raw video file using FFmpeg, apply the select video filter with a mathematical expression that evaluates to true for frames you wish to keep. This method allows precise control over frame selection without re-encoding, preserving the original quality while altering the temporal structure of the footage.

Understanding Frame Selection Logic

When working with raw video formats such as YUV420P or uncompressed AVI containers, the goal is often to modify the playback speed or correct duplication errors introduced during digitization. The core mechanism for this task is the select filter. Unlike cutting tools that rely on timestamps, select operates on individual frame indices, making it ideal for periodic removal tasks. The logic relies on modulo arithmetic. In mathematics, the modulo operation finds the remainder of a division. If we divide the current frame number by N, the result cycles from 0 to N-1. To remove every Nth frame, we must identify when the remainder equals zero. However, because frame indexing typically starts at 0, the "first" frame is index 0. Depending on your definition of "every Nth frame," you may need to adjust the offset. For most practical applications involving duplicated film frames, removing frames where the index modulo N equals zero is the standard approach. It is crucial to distinguish between keeping and dropping frames. A common error is assuming that a condition returning true drops the frame. In FFmpeg's select filter, expressions that evaluate to non-zero (true) keep the frame, while zero (false) drops it. Therefore, to remove every Nth frame, you must construct an expression that returns false only for those specific indices.

Constructing the Filter Expression

The syntax for the filter is -vf select='expr'. Inside the quotes, you define the expression. Let us assume you want to remove every 5th frame. The frame index is available via the variable n. To remove every 5th frame, you can use the following command:

ffmpeg -i input.yuv -vf "select=not(mod(n\,5))" -c:v rawvideo output_trimmed.yuv

Here, mod(n,5) calculates the remainder of the frame number divided by 5. When n is 5, 10, 15, etc., the remainder is 0. The not() function inverts this: not(0) becomes 1 (true), which would actually *keep* the frame if used directly in some contexts, but wait—let us clarify the boolean logic carefully. In FFmpeg, select keeps frames where the expression is non-zero. If you use select=not(mod(n,5)): - For n=0: mod(0,5)=0. not(0)=1. Frame kept. - For n=5: mod(5,5)=0. not(0)=1. Frame kept. This actually keeps every 5th frame and removes the others. This contradicts the goal of removing every Nth frame. To remove every Nth frame, you want to drop frames where mod(n,N) == 0. You should instead use:

ffmpeg -i input.yuv -vf "select=mod(n\,5)!=0" -c:v rawvideo output_trimmed.yuv

In this corrected version: - For n=0: mod(0,5)=0. 0!=0 is false (0). Frame dropped. - For n=1: mod(1,5)=1. 1!=0 is true (1). Frame kept. - For n=5: mod(5,5)=0. 0!=0 is false (0). Frame dropped. This ensures that frames at indices 0, 5, 10, etc., are excluded from the output. Note that raw video files often start counting from 0, so the very first frame might be removed depending on your specific needs. If you wish to keep the first frame but remove every 5th thereafter, you might adjust the logic to select=(n==0)+mod(n\,5)!=0, though simple modulo exclusion is usually sufficient for correcting duplicate frame artifacts.

Handling Raw Video Parameters

Raw video files do not contain header information regarding resolution, pixel format, or frame rate. FFmpeg cannot automatically determine these properties from a .yuv or .raw file. You must explicitly provide them using command-line flags before processing. Failure to do so will result in errors or corrupted output. You need to specify: - -s: The spatial resolution (e.g., 1920x1080). - -pix_fmt: The pixel format (e.g., yuv420p). - -r: The frame rate (e.g., 30 or 24000/1001). A complete command line looks like this:

ffmpeg -s 1920x1080 -pix_fmt yuv420p -r 30 -i input.yuv -vf "select=mod(n\,5)!=0" -c:v rawvideo -s 1920x1080 -pix_fmt yuv420p -r 30 output_trimmed.yuv

Notice that the output parameters also require -s, -pix_fmt, and -r. Since we are not re-encoding (-c:v rawvideo), FFmpeg writes the raw bytes directly. The frame rate in the output file remains the same as the input, but the actual duration of the video file will decrease because there are fewer frames. If you intend to play this back at the same perceived speed, you would need to increase the frame rate or use a player that handles variable frame rates, though standard players will simply play faster due to the missing frames. For more complex editing tasks, such as combining multiple segments after trimming, refer to our guide on how to concatenate videos with FFmpeg to understand how to merge the resulting clips seamlessly.

Alternative Approaches and Considerations

While select is powerful, it processes frames sequentially. For very long raw files, this can be slow because FFmpeg must decode every frame to evaluate the expression, even if it drops it. There is no "seek" optimization for select filters. If you are dealing with compressed formats like MP4 or MKV, consider using the fps filter if your goal is simply to change the frame rate rather than remove specific indices. However, fps averages or duplicates frames to match a target rate, whereas select gives exact control over which discrete frames survive. Another consideration is audio synchronization. If your raw video is embedded in a container with audio, removing video frames will desynchronize the audio track unless you also trim the audio to match the new frame count. For pure raw video streams without audio, this is not an issue. If you need to overlay images or text on the remaining frames after filtering, you can chain filters. For example, to add a border to only the kept frames, you could combine select with drawbox. See our article on using FFmpeg to overlay images on video frames for techniques on visual modifications.

Frequently Asked Questions

Does removing frames change the file size?

Yes, significantly. Since raw video files store every pixel unencoded, each frame has a fixed byte size. Removing every Nth frame reduces the total number of frames, thereby proportionally reducing the file size. For example, removing every 5th frame reduces the file size by approximately 20%.

Will the video play faster or slower?

The playback speed depends on how you view the file. If you play the trimmed raw file with the same frame rate setting as the original, it will play faster because there are fewer frames to display per second of real-time. To maintain the original perceived speed, you must either increase the reported frame rate or accept that the action is sped up.

Can I use this method on MP4 files?

Yes, but you must account for keyframes. Most video codecs only allow random access at keyframe boundaries. Using select on a compressed format may cause seeking issues or require full decoding. It is generally recommended for raw video or when using lossless codecs. For compressed video, consider transcoding with the filter applied if precise frame removal is critical.

What if I want to keep every Nth frame instead?

Simply invert the logic. Instead of mod(n\,N)!=0, use mod(n\,N)==0. This expression evaluates to true only for frames at indices 0, N, 2N, etc., keeping only those specific frames and dropping all others.

How do I handle the first frame being removed?

Since frame indexing starts at 0, mod(0, N) is always 0. If you use mod(n\,N)!=0, the first frame is dropped. To keep the first frame, modify the expression to select=(n==0)+mod(n\,N)!=0. This adds a condition that forces the expression to true for the first frame regardless of the modulo result.