You recorded a webinar or gameplay session and want clean chapters for each topic. The community\u2019s go-to tool for slicing video on any platform is FFmpeg. If you have ever scrolled through threads comparing flags and timestamp formats, you are not alone. <\/p>\n\n\n\n
Hi all,<\/em> I need a clear explanation of how to split a video with ffmpeg. Could someone explain the simplest command to cut a single clip between two timestamps, and also how to split a file into equal segments, like 10 seconds each? I\u2019d also appreciate tips on the differences between frame-accurate cuts and fast splitting without re-encoding, and any gotchas I should be aware of when it comes to audio tracks, subtitles, or output settings.<\/em> Great question! FFmpeg provides multiple ways to split video. The best method depends on whether you need fast, lossless cuts or frame-accurate results. Let\u2019s check them out.<\/p>\n\n\n\n Fast and lossless when your cut points are near keyframes:<\/p>\n\n\n For frame-accurate results, re-encode the video stream:<\/p>\n\n\n Use the segment muxer. This is ideal for chunking content into fixed durations:<\/p>\n\n\n If you have multiple chapters, loop through ranges and create clips:<\/p>\n\n\n Once you upload your video, you can generate trimmed segments on the fly with URL parameters or SDKs. For example, to deliver a 15 second clip from 12s to 27s you can apply a trim transformation using start and end offsets, then create multiple derived assets for each chapter as needed.<\/p>\n\n\n
Thanks!<\/em><\/p>\n\n\n\nAnswer:<\/h2>\n\n\n\n
1) Cut a single clip between timestamps<\/h3>\n\n\n\n
# Fast<\/span> split<\/span> -<\/span> copy<\/span> streams<\/span>, may<\/span> cut<\/span> at<\/span> nearest<\/span> keyframe<\/span> boundaries<\/span>\nffmpeg<\/span> -ss<\/span> 12 -t<\/span> 15 -i<\/span> input<\/span>.mp4<\/span> -c<\/span> copy<\/span> clip1<\/span>.mp4<\/span>\n# Starts<\/span> at<\/span> 12s<\/span> and<\/span> outputs<\/span> 15s<\/span> (ends<\/span> at<\/span> ~27s<\/span>)<\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n# Frame-accurate<\/span> split<\/span> -<\/span> re-encodes<\/span> video<\/span>\nffmpeg<\/span> -ss<\/span> 12 -t<\/span> 15 -i<\/span> input<\/span>.mp4<\/span> -c<\/span>:v<\/span> libx264<\/span> -crf<\/span> 18 -preset<\/span> veryfast<\/span> -c<\/span>:a<\/span> copy<\/span> clip1<\/span>.mp4<\/span><\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n\n
-ss <\/code>before -i <\/code>is faster. Using -ss<\/code> after -i<\/code> can be more accurate for stream copy, but is slower.<\/li>\n\n\n\n-t<\/code> (duration) instead of -to<\/code> to avoid confusion about absolute vs relative end times.<\/li>\n<\/ul>\n\n\n\n2) Split a video into equal segments<\/h3>\n\n\n\n
# Split<\/span> into<\/span> 10 second<\/span> parts<\/span> without<\/span> re-encoding<\/span>\nffmpeg<\/span> -i<\/span> input<\/span>.mp4<\/span> -c<\/span> copy<\/span> -map<\/span> 0 -f<\/span> segment<\/span> \\\n\u00a0 -segment_time<\/span> 10 -reset_timestamps<\/span> 1 parts_<\/span>%03d<\/span>.mp4<\/span><\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n\n
-reset_timestamps 1<\/code> restarts timestamps at zero for each segment.<\/li>\n\n\n\n# Re-encode<\/span> and<\/span> force<\/span> keyframes<\/span> every<\/span> 10s<\/span> for<\/span> clean<\/span> boundaries<\/span>\nffmpeg<\/span> -i<\/span> input<\/span>.mp4<\/span> -c<\/span>:v<\/span> libx264<\/span> -crf<\/span> 20 -preset<\/span> fast<\/span> -c<\/span>:a<\/span> copy<\/span> \\\n\u00a0 -force_key_frames<\/span> \"expr<\/span>:gte(t<\/span>,n_forced<\/span>*10)\" \\\n\u00a0 -f<\/span> segment<\/span> -segment_time<\/span> 10 -reset_timestamps<\/span> 1 parts_<\/span>%03d<\/span>.mp4<\/span><\/code><\/span>Code language:<\/span> CSS<\/span> (<\/span>css<\/span>)<\/span><\/small><\/pre>\n\n\n3) Split by a list of ranges<\/h3>\n\n\n\n
# Bash example - ranges are start,duration per line<\/span>\nwhile<\/span> IFS=, read -r start dur name; do<\/span>\n\u00a0 ffmpeg -ss \"$start\"<\/span> -t \"$dur\"<\/span> -i input.mp4 -c copy \"$name\"<\/span>.mp4\ndone < ranges.csv\n\n# ranges.csv<\/span>\n# 12,15,intro<\/span>\n# 27,28,feature_demo<\/span>\n# 55,20,q_and_a<\/span><\/code><\/span>Code language:<\/span> PHP<\/span> (<\/span>php<\/span>)<\/span><\/small><\/pre>\n\n\n4) Audio, subtitles, metadata, and fast start<\/h3>\n\n\n\n
\n
-map 0<\/code> to include all tracks, or select specific ones like -map 0:v:0 -map 0:a:1<\/code>.<\/li>\n\n\n\n-sn<\/code>.<\/li>\n\n\n\n-movflags +faststart<\/code> to move the index to the front.<\/li>\n\n\n\nDo the same with Cloudinary (URL-based trimming and automation)<\/h3>\n\n\n\n
\/\/ Node.js example - create derived clips using start\/end offsets<\/span>\nimport<\/span> { v2 as<\/span> cloudinary } from<\/span> 'cloudinary'<\/span>;\n\ncloudinary.uploader.explicit('videos\/my_talk'<\/span>, {\n\u00a0 resource_type<\/span>: 'video'<\/span>,\n\u00a0 eager<\/span>: [\n\u00a0 \u00a0 { transformation<\/span>: [{ so<\/span>: 12<\/span>, eo<\/span>: 27<\/span> }], format<\/span>: 'mp4'<\/span> }, \u00a0 \/\/ intro<\/span>\n\u00a0 \u00a0 { transformation<\/span>: [{ so<\/span>: 27<\/span>, eo<\/span>: 55<\/span> }], format<\/span>: 'mp4'<\/span> }, \u00a0 \/\/ demo<\/span>\n\u00a0 \u00a0 { transformation<\/span>: [{ so<\/span>: 55<\/span>, eo<\/span>: 75<\/span> }], format<\/span>: 'mp4'<\/span> }\u00a0 \u00a0 \/\/ Q&A<\/span>\n\u00a0 ],\n\u00a0 eager_async<\/span>: false<\/span>\n});<\/code><\/span>Code language:<\/span> JavaScript<\/span> (<\/span>javascript<\/span>)<\/span><\/small><\/pre>\n\n\n