FFmpeg Trim Without Re-encoding: Keyframe Boundaries and Stream Copy

Why Re-encoding Should Be Avoided

Every time video is re-encoded, quality drops slightly and time is spent computing the new encoded stream. For trimming operations (cutting the start, end, or middle of a video), re-encoding is often unnecessary.

FFmpeg's -c copy flag enables "stream copy" mode: the encoded video and audio data are copied bit-perfect from input to output. The result is a trimmed file in seconds, not minutes, with zero quality loss.

The catch: stream copy only works at keyframe boundaries. Most modern video has keyframes every 2-10 seconds. Trimming between keyframes either:

1. Loses frames at the start (until the next keyframe)
2. Includes extra frames at the start (back to the previous keyframe)

This post covers the practical FFmpeg trim workflow with both stream copy and frame-accurate trimming. For broader FFmpeg context, see HEVC to H.264 for Premiere.

Stream Copy Trim

The basic command:

ffmpeg -ss 00:00:30 -to 00:05:00 -i input.mp4 -c copy output.mp4

Parameters:

• -ss 00:00:30: start at 30 seconds
• -to 00:05:00: end at 5 minutes
• -c copy: stream copy (no re-encoding)
• -i input.mp4: input file

The output is a 4:30 clip from the source. Processing time: 1-2 seconds for a 1-hour source. Quality: identical to source.

Keyframe Boundary Effect

When the trim points fall on keyframes, output is exactly as specified. When they don't:

For most modern video (2-second GOP), the start time is within ±2 seconds of requested. For exact start: re-encode or use a different approach.

Frame-Accurate Trim with Re-encoding

For frame-accurate trim:

ffmpeg -ss 00:00:30.5 -to 00:05:00 -i input.mp4 \
  -c:v libx264 -preset ultrafast -crf 18 \
  -c:a aac -b:a 192k \
  -movflags +faststart \
  output.mp4

CRF 18 is visually lossless. The -preset ultrafast minimizes re-encoding time at slight file size cost. Output is exactly 30.5 seconds to 5:00 from source.

For maximum quality at the cost of more time: -preset slow -crf 18.

Hybrid Approach

For best of both worlds: re-encode the start and end, stream-copy the middle:

# Stream copy from nearest keyframe at 30s to 5:00
ffmpeg -ss 00:00:30 -to 00:05:00 -i input.mp4 -c copy temp_middle.mp4

# Re-encode just the part before 30s
ffmpeg -ss 00:00:28 -to 00:00:30 -i input.mp4 \
  -c:v libx264 -preset ultrafast -crf 18 \
  -c:a aac \
  temp_start.mp4

# Concat
ffmpeg -f concat -safe 0 -i <(printf "file '%s'\nfile '%s'\n" "$(pwd)/temp_start.mp4" "$(pwd)/temp_middle.mp4") \
  -c copy output.mp4

This produces frame-accurate trim at minimal re-encoding cost. Most of the file is stream copy; only the few seconds at the boundaries are re-encoded.

Multiple Trim Operations

For trimming out a segment in the middle (split + concat):

# Get part before bad section
ffmpeg -i input.mp4 -ss 00:00:00 -to 00:02:00 -c copy part1.mp4

# Get part after bad section
ffmpeg -i input.mp4 -ss 00:03:00 -to 00:60:00 -c copy part2.mp4

# Concat
ffmpeg -f concat -safe 0 -i <(printf "file '%s'\nfile '%s'\n" "$(pwd)/part1.mp4" "$(pwd)/part2.mp4") \
  -c copy output.mp4

The concat preserves quality (no re-encode). Can be combined with the hybrid approach for frame-accurate cuts.

For batch trim patterns, see Batch Processing Files Guide.

Identifying Keyframes

To check where keyframes are in your video:

ffprobe -v error -select_streams v:0 \
  -show_entries packet=pts_time,flags \
  -of csv input.mp4 | grep K

Output is a list of timestamps where keyframes occur. Use these for choosing trim points that work with stream copy.

For a video with 2-second GOP at 30 fps: keyframes at every 60 frames (2 seconds).

When Re-encoding Is Necessary

You can't avoid re-encoding when:

• Trimming requires frame-accurate boundaries
• Source has very long GOP (10+ seconds)
• Combining multiple sources with different codecs
• Adding effects, filters, or overlays
• Format conversion (MP4 to WebM, etc.)
• Resolution change

For these: use re-encode workflow. CRF 18 visually lossless preserves quality.

Audio Trim

For audio file trimming:

ffmpeg -ss 00:00:30 -to 00:05:00 -i input.mp3 -c copy output.mp3

MP3 has very small frames (typically 26ms each), so trim is essentially frame-accurate without re-encoding. Same applies to AAC.

For lossless audio (FLAC, WAV):

ffmpeg -ss 00:00:30 -to 00:05:00 -i input.flac -c copy output.flac

Stream copy with audio is much more accurate than with video.

Common Issues

Output starts before requested time: keyframe boundary issue. Either accept the offset, or re-encode for frame accuracy.

Output has black frames at start: source has GOP and the player decodes from the keyframe. Some players handle this gracefully; others show black until next keyframe.

Audio out of sync after trim: audio offset from video keyframes. Re-mux with -vsync cfr.

File size larger than expected: stream copy preserves source bitrate; can't reduce file size with -c copy. Re-encode at higher CRF for smaller files.

FFmpeg trim ignores negative offset: -ss doesn't accept negative values. Use -itsoffset for input offset adjustments.

For broader video processing context, see HEVC to H.264 for Premiere.

Trim with Subtitle Tracks

If your input has subtitles:

ffmpeg -ss 00:00:30 -to 00:05:00 -i input.mkv \
  -map 0:v -map 0:a -map 0:s \
  -c copy \
  output.mkv

The -map flags include video, audio, and subtitle streams. Stream copy preserves them.

For subtitle conversion details, see SRT vs VTT vs ASS.

Trim with Multiple Audio Tracks

For multi-language video:

ffmpeg -ss 00:00:30 -to 00:05:00 -i input.mkv \
  -map 0:v -map 0:a:0 -map 0:a:1 -map 0:a:2 \
  -c copy \
  output.mkv

Maps the video and the first three audio tracks. Useful for Blu-ray rips with multiple language tracks.

For MKV vs MP4 context, see MKV vs MP4 Container.

Frequently Asked Questions

Should I always use stream copy?

For trim-only operations: yes, when accuracy permits. For complex edits: re-encode.

What's the speed difference?

Stream copy: 30x realtime typical. Re-encode: 0.5-1x realtime depending on settings.

Does stream copy work for all formats?

Most modern formats (MP4, MKV, WebM, MOV, FLV). Some legacy formats (AVI, MPEG-1) may have issues. Try stream copy first; if it fails, re-encode.

Can I trim with audio fade in/out?

Stream copy doesn't support audio effects. For fade in/out: re-encode the trim with audio filter:

ffmpeg -ss 00:00:30 -to 00:05:00 -i input.mp4 \
  -c:v libx264 -crf 18 \
  -c:a aac -af "afade=in:st=0:d=2,afade=out:st=268:d=2" \
  output.mp4

What about precise frame-by-frame edits?

For frame-accurate cuts every frame: re-encode. Stream copy is keyframe-aligned only.

How do I trim multiple segments?

Use the concat approach above. Multiple stream-copy trims, concat with -f concat filter.

Related Reading

• HEVC to H.264 for Premiere
• Batch Processing Files Guide
• Compress Video for WhatsApp

Bottom Line

For FFmpeg trimming without re-encoding: use -c copy flag, accept keyframe-boundary alignment. For frame-accurate trim: re-encode at CRF 18 with ultrafast preset. Hybrid approach for production work where most of the video is stream-copy and only edge frames re-encode. Our video compressor handles re-encoding at known-good settings.