FFmpeg Recipies

Any technology, no matter how primitive, is magic to those who don’t understand it. - Arthur C. Clarke / Mark Stanley

Here are some magical incantations that let me do help me do cool stuff with FFmpeg.

Fade vision + audio between 2 videos

ffmpeg -i big_buck_bunny.mp4 -i blackout.mkv -filter_complex \
"[0:v]fade=t=out:st=5:d=1:alpha=1,setpts=PTS-STARTPTS[v0];
[1:v]fade=t=in:st=0:d=1:alpha=1,setpts=PTS-STARTPTS+5/TB[v1];
[v0][v1]overlay[v];
[0][1]acrossfade=d=1[a]" \
-map [v] -map [a] result.mp4

• st=5:d=1 start fade out of first clip at 5 seconds, fade duration 1 second
• acrossfade=d=1 fade audio from clip 1 to clip 2, duration 1 second
• PTS-STARTPTS+5/TB the second clip needs to start 5 seconds after the first clip 5/TB is “5 divided by time base”

Thanks to this post: https://opensourceforu.com/2015/04/get-friendly-with-ffmpeg/

Mux mp4 + subtitles into an MKV

ffmpeg -fflags +genpts -i infile.mp4 -f srt -i subtitles.srt \
-map 0:0 -map 0:1 -map 1:0 -c:v copy -c:a copy -c:s srt outfile.mkv

-fflags +genpts necessary if you get Can't write packet with unknown timestamp errors.

Extract part of video using start + end times

Transcode to lossless Flac and while applying peak normalisation to audio track

#!/bin/bash
#
# process tab delmited 'runsheet' file with format
#	output name	input name	start time	end time
#
# where start/end time are hh:mm:ss
#

tosecs() {
   date '+%s' --date="$1"
}

fromsecs() {
 ((h=${1}/3600))
 ((m=(${1}%3600)/60))
 ((s=${1}%60))
 printf "%02d:%02d:%02d\n" $h $m $s
}

IFS=$'\n'
for i in `cat runsheet`; do

	oname=`echo $i | cut -f1`
	iname=`echo $i | cut -f2 | tr -d "'"`
	start=`echo $i | cut -f3`
	end=`echo $i | cut -f4`

	to=""
	if [[ -n $end ]]; then
		t=$(( $(tosecs "$end") - $(tosecs "$start") ))
		to="-t $(fromsecs $t)"
	fi
	if [[ -n $start ]]; then
		ss="-ss $start"
	else
		ss="-ss 00:00:00"
	fi

	cmd="ffmpeg $ss $to -i \"../$iname\" -vcodec copy -acodec flac -filter:a loudnorm $oname.mkv"
	echo $cmd
	eval $cmd

	date
	echo
done

Record from microphone, encode to Opus and send to network socket

ffmpeg -f pulse -i default -acodec libopus -b:a 96000 -vbr on -compression_level 10 -f rtp rtp://127.0.0.1:1234

Play from RTP stream:

Create SDP file using details output from ffmpeg command.

v=0
o=- 0 0 IN IP4 127.0.0.1
s=No Name
c=IN IP4 127.0.0.1
t=0 0
a=tool:libavformat 58.12.100
m=audio 1234 RTP/AVP 97
b=AS:96
a=rtpmap:97 opus/48000/2
a=fmtp:97 sprop-stereo=1

Then pass SDP file to RTP client:

ffplay -i opus.sdp -protocol_whitelist file,udp,rtp

Video Capture Using EasyCAP USB analog to digital convertor

• Capture audio from device1 (hw:1)
• Capture video from /dev/video0 as PAL (720x576 50hz)
• de-interlace (yadif)
• encode video using ‘fast’ preset (using slow was getting dropped frames)
• encode audio as AAC 128kb

software encoding

ffmpeg \
  -f alsa -ac 2 -thread_queue_size 512 -i hw:1 \
  -f v4l2 -standard PAL -thread_queue_size 512 -i /dev/video0 \
  -vf yadif -c:v libx264 -preset fast -crf 23 \
  -c:a aac -b:a 128k \
  output.mkv

hardware encoding using VAAPI

ffmpeg \
  -f alsa -ac 2 -thread_queue_size 1024 -i hw:1 \
  -f v4l2 -standard PAL -thread_queue_size 1024 -i /dev/video0 \
  -vaapi_device /dev/dri/renderD128 -vf 'format=nv12,hwupload' -threads 4 -vcodec h264_vaapi -qp:v 23 \
  -c:a aac -b:a 128k \
  output.mkv

No deinterlacing with hardware encoding :(

Convert sequence of JPEG images to MP4 video

Simple glob:

ffmpeg -r 24 -i '*.JPG' -s hd1080 -vcodec libx264 timelapse.mp4

Start from DSC_0079.JPG

ffmpeg -r 24 -f image2 -start_number 79 -i DSC_%04d.JPG -s hd1080 -vcodec libx264 timelapse2.mp4

• -r 24 - output frame rate
• -s hd1080 - 1920x1080 resolution

Slower, better quality

Add the following after -vcodec libx264 to achieve better quality output

-crf 18 -preset slow

Bulk convert JPGs to 1920x1080, centered

convert input.jpg -resize '1920x1080^' -gravity center -crop '1920x1080+0+0' output.jpg