基于FFmpeg实现MP4转WebM及时长自适应的10秒剪辑生成

Fixed Timestamp Solution #

Let’s start with the exact scenario you outlined: pulling 4 x 2.5-second clips from specific time points in your 12-minute input.mp4 and merging them into output.webm.

Option 1: Single Command (No Temporary Files) #

You can handle both extraction and merging in one go using FFmpeg’s concat filter. Here’s the full command:

ffmpeg -i input.mp4 \
-filter_complex "\
[0:v]trim=start=150:end=152.5,setpts=PTS-STARTPTS[v0]; \
[0:a]atrim=start=150:end=152.5,asetpts=PTS-STARTPTS[a0]; \
[0:v]trim=start=300:end=302.5,setpts=PTS-STARTPTS[v1]; \
[0:a]atrim=start=300:end=302.5,asetpts=PTS-STARTPTS[a1]; \
[0:v]trim=start=450:end=452.5,setpts=PTS-STARTPTS[v2]; \
[0:a]atrim=start=450:end=452.5,asetpts=PTS-STARTPTS[a2]; \
[0:v]trim=start=600:end=602.5,setpts=PTS-STARTPTS[v3]; \
[0:a]atrim=start=600:end=602.5,asetpts=PTS-STARTPTS[a3]; \
[v0][a0][v1][a1][v2][a2][v3][a3]concat=n=4:v=1:a=1[v][a]" \
-map "[v]" -map "[a]" output.webm

Quick breakdown:

• trim/atrim isolate the video/audio segments (times converted to seconds: 2.5min = 150s, end time = start + 2.5s)
• setpts/asetpts reset timestamps so each clip starts at 0 (required for smooth concatenation)
• The concat filter combines all 4 clips into a single stream

Option 2: Modular Workflow (Easier Debugging) #

If you prefer step-by-step control, extract clips first then merge:

1. Pull individual clips:

ffmpeg -ss 150 -i input.mp4 -t 2.5 -c copy clip1.webm
ffmpeg -ss 300 -i input.mp4 -t 2.5 -c copy clip2.webm
ffmpeg -ss 450 -i input.mp4 -t 2.5 -c copy clip3.webm
ffmpeg -ss 600 -i input.mp4 -t 2.5 -c copy clip4.webm

2. Create a clips.txt file listing your clips:

file 'clip1.webm'
file 'clip2.webm'
file 'clip3.webm'
file 'clip4.webm'

3. Merge with the concat demuxer:

ffmpeg -f concat -safe 0 -i clips.txt -c copy output.webm

Note: -c copy copies streams directly (fast), but if your input codecs aren’t WebM-compatible, remove -c copy to let FFmpeg re-encode (or specify -c:v libvpx-vp9 -c:a opus for optimal WebM quality).

Auto-Adjusting Timestamps Based on Input Duration #

Absolutely! You can dynamically calculate start times using ffprobe to get the video’s total duration, then use a script to spread the clips evenly across the video.

Bash Script Example #

This script automatically spaces 4 clips across any input video, avoiding the very start/end to prevent partial clips:

#!/bin/bash

INPUT="input.mp4"
OUTPUT="output.webm"
CLIP_DURATION=2.5
NUM_CLIPS=4

# Get total video duration in seconds (rounded to 2 decimals)
TOTAL_DURATION=$(ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "$INPUT" | awk '{printf "%.2f", $0}')

# Calculate even spacing with buffer at start/end
BUFFER=$(echo "$CLIP_DURATION / 2" | bc -l)
AVAILABLE_DURATION=$(echo "$TOTAL_DURATION - $CLIP_DURATION" | bc)
INTERVAL=$(echo "$AVAILABLE_DURATION / ($NUM_CLIPS - 1)" | bc -l)

# Build filter complex string
FILTER_COMPLEX=""
for i in $(seq 0 $((NUM_CLIPS - 1))); do
    START_TIME=$(echo "$BUFFER + ($i * $INTERVAL)" | bc -l)
    END_TIME=$(echo "$START_TIME + $CLIP_DURATION" | bc -l)
    FILTER_COMPLEX+="[0:v]trim=start=$START_TIME:end=$END_TIME,setpts=PTS-STARTPTS[v$i]; "
    FILTER_COMPLEX+="[0:a]atrim=start=$START_TIME:end=$END_TIME,asetpts=PTS-STARTPTS[a$i]; "
done

# Add concat filter
FILTER_COMPLEX+=$(printf "[v%d][a%d]" $(seq 0 $((NUM_CLIPS - 1))) | tr '\n' ' ')
FILTER_COMPLEX+="concat=n=$NUM_CLIPS:v=1:a=1[v][a]"

# Run FFmpeg
ffmpeg -i "$INPUT" -filter_complex "$FILTER_COMPLEX" -map "[v]" -map "[a]" "$OUTPUT"

How it works:

1. ffprobe fetches the exact duration of your input video
2. Calculates start times so clips are evenly distributed (with a small buffer to avoid cutting off at the edges)
3. Dynamically builds the filter complex and runs FFmpeg

Customization Tips #

• To remove the buffer (start clips right at the beginning), set BUFFER=0 and adjust AVAILABLE_DURATION to $TOTAL_DURATION - $CLIP_DURATION
• For Windows, adapt this using PowerShell or WSL
• Adjust NUM_CLIPS or CLIP_DURATION to fit your needs

内容的提问来源于stack exchange，提问作者EmJeiEn