Video Technology

#!/bin/bash VIDSOURCE="$1" RESOLUTION="854x480" BITRATE1="800000" BITRATE2="600000" BITRATE3="400000" AUDIO_OPTS="-c:a libfaac -b:a 160000 -ac 2" VIDEO_OPTS1="-s $RESOLUTION -c:v libx264 -b:v $BITRATE1 -vprofile

baseline -preset medium -x264opts level=41"

VIDEO_OPTS2="-s $RESOLUTION -c:v libx264 -b:v $BITRATE2 -vprofile

baseline -preset medium -x264opts level=41"

VIDEO_OPTS3="-s $RESOLUTION -c:v libx264 -b:v $BITRATE3 -vprofile

baseline -preset medium -x264opts level=41"

OUTPUT_HLS="-hls_time 3 -hls_list_size 10 -hls_wrap 30 -start_number 1" ffmpeg -i "$VIDSOURCE" -y -threads 4 \        $AUDIO_OPTS $VIDEO_OPTS1 $OUTPUT_HLS stream_hi.m3u8 \        $AUDIO_OPTS $VIDEO_OPTS2 $OUTPUT_HLS stream_med.m3u8 \        $AUDIO_OPTS $VIDEO_OPTS3 $OUTPUT_HLS stream_low.m3u8

For starters here is the command to create a stream that is GOP aligned, that is has keyframes at regular intervals that can be nicely segmented. It is recommended to pull down and compile the latest version of ffmpeg for this as some of the distros can be a bit out of date.

ffmpeg -y -i %(filename)s -pix_fmt yuv420p -vcodec libx264 -acodec libfaac -r %(fps)s -profile:v baseline -b:v %(bitrate)sk -maxrate %(bitrate)sk  -force_key_frames %(keyframe_str)s -s %(width)sx%(height)s %(target_dir)s%(newFile)s

The s%(filename) are in Python format where the %s(var) bit gets replaced with your value.

For example:

ffmpeg -y -i infile.mp4 -pix_fmt yuv420p -vcodec libx264 -acodec libfaac -r 25 -profile:v baseline -b:v 2000k -maxrate 2500k  -force_key_frames 50s -s 640x360 /tmp/newfile.ts

I like to do segmenting as a second step just to keep the whole process a bit cleaner, especially if you want to produce multiple formats off the same initial encode:

ffmpeg -y -i %(target_dir)s%(filename)s -codec copy -map 0 -f segment -segment_list %(target_dir)sindex_%(bitrate)s.m3u8 -segment_time %(segment_size)s  -segment_list_type m3u8 %(filenameNoExt)s_%(count)s.ts

Or you could wrap it all into one command e.g.

ffmpeg -y -i %(filename)s -pix_fmt yuv420p -vcodec libx264 -acodec libfaac -r %(fps)s -profile:v baseline -b:v %(bitrate)sk -maxrate %(bitrate)sk  -force_key_frames %(keyframe_str)s -s %(width)sx%(height)s %-f segment -segment_list %(target_dir)sindex_%(bitrate)s.m3u8 -segment_time %(segment_size)s  -segment_list_type m3u8 %(filenameNoExt)s_%(count)s.ts

Working example:

ffmpeg -y -i infile.mp4 -pix_fmt yuv420p -vcodec libx264 -acodec libfaac -r 25 -profile:v baseline -b:v 1500k -maxrate 2000k -force_key_frames 50 -s 640x360 -map 0 -flags -global_header -f segment -segment_list /tmp/index_1500.m3u8 -segment_time 10 -segment_format mpeg_ts -segment_list_type m3u8 /tmp/segment%05d.ts

Here is an example of capturing from the desktop Linux and streaming straight to a local web server configured for low latency:

ffmpeg -f x11grab -s `xdpyinfo | grep 'dimensions:'|awk '{print $2}'` -r 25 -i :0.0 -pix_fmt yuv420p -vcodec libx264 -acodec libfaac -r 25 -profile:v baseline -b:v 1500k -x264opts keyint=25 -s 640x360 -map 0 -flags -global_header -f segment -segment_list index_1500.m3u8 -segment_time 1 -segment_format mpeg_ts -segment_list_type m3u8 -segment_list_flags +live -segment_list_size 2 segment%05d.ts

Tested with:

ffmpeg version N-51554-g1c0d8f2 Copyright (c) 2000-2013 the FFmpeg developers

  built on Apr  7 2013 18:35:14 with llvm-gcc 4.2.1 (LLVM build 2336.11.00)

  configuration: --enable-libx264 --enable-libfaac --enable-nonfree --enable-gpl

  libavutil      52. 24.100 / 52. 24.100

  libavcodec     55.  2.100 / 55.  2.100

  libavformat    55.  1.100 / 55.  1.100

  libavdevice    55.  0.100 / 55.  0.100

  libavfilter     3. 49.100 /  3. 49.100

  libswscale      2.  2.100 /  2.  2.100

  libswresample   0. 17.102 /  0. 17.102

  libpostproc    52.  2.100 / 52.  2.100

Android suffers from not having a good across the board player, especially for HLS. There appear to be a few approaches:

Chance it with the native player

Link through to a VLC variant using Android NDK/JNI

Link through to a ffmpeg variant using Android NDK/JNI

Run a HLS proxy that converts the HLS to RTSP and then sends RTSP to the native player (SOL HLS does this)

NXP Cinexplayer

Sol HLS Player

Looked promising as it is supposed to just pickup the mime-type and then control the playback but had trouble getting it to pick up the stream.

Note that if you want to like to a live HLS stream use the intent httplive:// works very well and even with a token protected Akamai stream.

After more testing am less convinced on this one as it takes the stream and then does a live transmux from HLS to RTSP and pipes that to the local player. The end result can be quite a lot of pixelisation.

Ooyala Hook

Only works with the Ooyala backend and appears to be based on the VisualOn player.

VisualOn

NXPlayer

ffmpeg/libav variant

MoboPlayer

Nexstreaming

Vitamio

Hoperun

http://www.hoperun.com/hoperun/

ffmpeg/libav

For starters do the following in the AWS console:

First create a new server e.g. use the Default 64 bit Amazon Linux instance

Create a new security group when you create the server

To the 

Create an Elastic IP and associate with the instance

First install vsftpd: sudo apt-get install vsftpd

Edit /etc/vsftpd.conf and make sure you have the following settings

anonymous_enable=no

local_enable=yes

write_enable=yes

chroot_local_user=yes

userlist_file=/etc/vsftpd.userlist

userlist_enable=YES

userlist_deny=NO

pasv_max_port=41000

pasv_min_port=40000

port_enable=YES

pasv_enable=YES

pasv_address=<Elastic IP address>

Adding a user:

Create a home dir: sudo mkdir /home/ftpuser

Create the user account: sudo useradd -d /home/ftpuser

Set the users passwd: sudo passwd ftpuser

Change owner of the dir: sudo chown ftpuser /home/ftpuser

Add the user to the allowed ftp user list: sudo vi /etc/vsftpd.userlist

ffmpeg -i sbs.mp4 -vf mp=stereo3d -acodec copy -threads 10 -b:v 10000k -preset ultrafast -vcodec libx264 ~/test2.mkv

http://delogics.blogspot.it/2012/01/sbspolarizedstereo-3d-to-alaglyphred.html

And more good info this time using mencoder:

ffprobe is quite a good little tool for evaluating keyframes in a video and here is a simple statement that can show you an easy way to view the GOP structure of a video:

ffprobe -show_frames out_gop12.ts |grep 'media_type=video\|pict_type\|coded_picture_number'|less

Now on doing this I am seeing that ffmpeg and x264 aren't necessarily following the GOP structure that I have set which has a max keyframe distance of 12.

Sample output:

media_type=video

pict_type=I

coded_picture_number=0

media_type=video

pict_type=P

coded_picture_number=1

media_type=video

pict_type=I

coded_picture_number=2

media_type=video

pict_type=P

coded_picture_number=3

media_type=video

pict_type=P

coded_picture_number=4

media_type=video

pict_type=P

coded_picture_number=5

media_type=video

pict_type=B

coded_picture_number=7

media_type=video

pict_type=P

coded_picture_number=6

media_type=video

pict_type=B

coded_picture_number=10

media_type=video

pict_type=B

coded_picture_number=9

media_type=video

pict_type=B

coded_picture_number=11

media_type=video

pict_type=P

coded_picture_number=8

media_type=video

pict_type=B

coded_picture_number=14

Here is a good post with some more info on the topic

http://forums.creativecow.net/thread/291/71

I have also created a little Python script that will process the output to show you when keyframes occur:

#!/usr/bin/python

import os

import re

import sys

import subprocess

#infile = sys.argv[1]

infile= "/tmp/ffmpeg_main_profile/out_gop12.ts"

lines_past_type = 20

line_count = 0

print_line = False

gop = 12

frame = 0

found_iframe = False

cmd = "ffprobe -show_frames %s" % infile

# -print_format json

process = subprocess.Popen(cmd,shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE \

                               ,cwd=target_dir)

while True:

  line = process.stdout.readline()

  if line != '':

    #the real code does filtering here

    #print line.rstrip()

    if (line.rstrip()=="media_type=video"):

        print_line = True

    if (print_line and (line_count < lines_past_type)):

                     #print line.rstrip()

        if re.match(r'pict_type=I.*',line):

            found_iframe = True

            #print line.rstrip()

        match = re.match(r'coded_picture_number=(.*)',line)

        if (match):

            frame =  match.group(1)

            if (found_iframe):

                print "I Frame: %s" % frame

                div12 = float(frame) / 12

                found_iframe = False

                 line_count+=1

        if (line_count >= lines_past_type):

            print_line = False

            line_count = 0

  else:

FFmpeg Howto

Table of Contents

Encoding:

H.264 Long GOP

D10 (aka Sony IMX)

XDCAM HD 50Mbps in QuickTime (for Final Cut Pro import)

DVCAM / DVCPRO25 / DVCPRO50

VC-3 (aka Avid DNxHD)

MPEG-2 I-frame only in Highest Quality

MPEG-2 Long GOP

MJPEG in Highest Quality

Muxing and Wrapping:

D10 into QuickTime (for Final Cut Pro import)

MPEG-2 Program Stream

Demuxing and Unwrapping:

MPEG-2 Program Stream

Timecode Management:

MPEG-2 Start Timecode

Misc:

Audio Volume Modification

Input Stream Selection

Sub-clip Creation

Make a Video File from a Single Frame

H.264 Long GOP Encoding

1-pass encoding:  ffmpeg -i <input_file> -vcodec libx264 -b <video_bitrate> -g <gop_size> -bf 3 -b_strategy 1 -coder 1 -qmin 10 -qmax 51 -sc_threshold 40 -flags +loop -cmp +chroma -me_range 16 -me_method hex -subq 5 -i_qfactor 0.71 -qcomp 0.6 -qdiff 4 -directpred 1 -flags2 +fastpskip -dts_delta_threshold 1 -acodec libfaac -ab <audio_bitrate> output.mp4  2-pass encoding:  ffmpeg -y -i <input_file> -pass 1 -vcodec libx264 -b <video_bitrate> -g <gop_size> -bf 3 -refs 6 -b_strategy 1 -coder 1 -qmin 10 -qmax 51 -sc_threshold 40 -flags +loop -cmp +chroma -me_range 16 -me_method umh -subq 7 -i_qfactor 0.71 -qcomp 0.6 -qdiff 4 -directpred 3 -flags2 +dct8x8+wpred+bpyramid+mixed_refs -trellis 1 -partitions +parti8x8+parti4x4+partp8x8+partp4x4+partb8x8 -acodec libfaac -ab <audio_bitrate> output.mp4  ffmpeg -y -i <input_file> -pass 2 -vcodec libx264 -b <video_bitrate> -g <gop_size> -bf 3 -refs 6 -b_strategy 1 -coder 1 -qmin 10 -qmax 51 -sc_threshold 40 -flags +loop -cmp +chroma -me_range 16 -me_method umh -subq 7 -i_qfactor 0.71 -qcomp 0.6 -qdiff 4 -directpred 3 -flags2 +dct8x8+wpred+bpyramid+mixed_refs -trellis 1 -partitions +parti8x8+parti4x4+partp8x8+partp4x4+partb8x8 -acodec libfaac -ab <audio_bitrate> output.mp4 

D10 (aka Sony IMX) Encoding

ffmpeg -i <input_file> -vcodec mpeg2video -r 25 -pix_fmt yuv422p -minrate 50000k -maxrate 50000k -b 50000k -intra -flags +ildct+low_delay -dc 10 -flags2 +ivlc+non_linear_q -ps 1 -qmin 1 -qmax 3 -top 1 -bufsize 2000000 -rc_init_occupancy 2000000 -rc_buf_aggressivity 0.25 -an output.m2v  Notes :  - Only works for 25 fps content.  - Change minrate, maxrate and b values to 30000k / 40000k / 50000k in order to produce 30/40/50 Mbps D10 essence.  - Set bufsize and rc_init_occupancy values to 1200000 / 1600000 / 2000000 for 30/40/50 Mbps D10 essence.  - Set qmax value to 8 / 3 / 3 for 30/40/50 Mbps D10 essence.  - Add the -padtop 32 option if you want to produce a 720x608 D10 image size from a 720x576 source. 

XDCAM HD 50Mbps in QuickTime (for Final Cut Pro import)

ffmpeg -i <input_file> -pix_fmt yuv422p -vcodec mpeg2video -flags +ildct+ilme -top 1 -dc 10 -flags2 +ivlc+non_linear_q -qmin 1 -lmin '1*QP2LAMBDA' -vtag xd5c -rc_max_vbv_use 1 -rc_min_vbv_use 1 -b 50000k -minrate 50000k -maxrate 50000k -bufsize 36408333 -bf 2 -aspect 16:9 -acodec pcm_s16be -f mov output.mov -newaudio 

DVCAM / DVCPRO25 / DVCPRO50 Encoding

ffmpeg -i <input_file> -pix_fmt yuv420p output_DVCAM.dv  ffmpeg -i <input_file> -pix_fmt yuv411p output_DVCPRO25.dv  ffmpeg -i <input_file> -pix_fmt yuv422p output_DVCPRO50.dv 

VC-3 (aka Avid DNxHD) Encoding

ffmpeg -i <input_file> -vcodec dnxhd -b <bitrate> -an output.mov  Notes :  - <bitrate> can take the following values : 36M, 60M, 90M, 120M, 185M ( please refer to the following table ).  - Add the following option for interlaced modes : -flags +ildct  - Add the following option for best quality mode ( very slow ! ) : -mbd rd  Supported Resolutions :  Project FormatResolutionFrame SizeBitsFPS<bitrate>1080i / 59.94DNxHD 2201920 x 1080829.97220M1080i / 59.94DNxHD 1451920 x 1080829.97145M1080i / 50DNxHD 1851920 x 1080825185M1080i / 50DNxHD 1201920 x 1080825120M1080p / 25DNxHD 1851920 x 1080825185M1080p / 25DNxHD 1201920 x 1080825120M1080p / 25DNxHD 361920 x 108082536M1080p / 24DNxHD 1751920 x 1080824175M1080p / 24DNxHD 1151920 x 1080824115M1080p / 24DNxHD 361920 x 108082436M1080p / 23.976DNxHD 1751920 x 1080823.976175M1080p / 23.976DNxHD 1151920 x 1080823.976115M1080p / 23.976DNxHD 361920 x 1080823.97636M1080p / 29.7DNxHD 451920 x 1080829.9745M720p / 59.94DNxHD 2201280x720859.94220M720p / 59.94DNxHD 1451280x720859.94145M720p / 50DNxHD 1751280x720850175M720p / 50DNxHD 1151280x720850115M720p / 23.976DNxHD 901280x720823.97690M720p / 23.976DNxHD 601280x720823.97660M

MPEG-2 I-frame only Highest Quality Encoding

ffmpeg -i <input_file> -vcodec mpeg2video -pix_fmt yuv422p -qscale 1 -qmin 1 -intra -an output.m2v 

MPEG-2 Long GOP Encoding

ffmpeg -i <input_file> -vcodec mpeg2video -b <video_bitrate> -g <gop_size> -bf 2 -b_strategy 1 -acodec mp2 -ab <audio_bitrate> -f vob output.mpg 

MJPEG in Highest Quality

ffmpeg -i <input_file> -vcodec mjpeg -qscale 1 -an output.avi 

D10 into QuickTime Wrapping

ffmpeg -i <input_file> -vcodec copy -acodec pcm_s16le -vtag mx5p -vbsf imxdump output.mov  Notes :  - This feature needs a patched version of FFmpeg not available in the current SVN, please contact Baptiste Coudurier for more information.  - Set -vtag to mx3p / mx4p / mx5p for D10 30/40/50 PAL or mx3n / mx4n / mx5n for D10 30/40/50 NTSC.  - Input D10 image size must be 720x608 ( use -padtop 32 option during D10 encoding if the original image size is 720x576 ).  - Final Cut Pro compliant. 

MPEG-2 Program Stream Muxing

ffmpeg -genpts 1 -i ES_Video.m2v -i ES_Audio.mp2 -vcodec copy -acodec copy -f vob output.mpg  Note : In order to mux multiple audio tracks into the same file :  ffmpeg -genpts 1 -i ES_Video.m2v -i ES_Audio1.mp2 -i ES_Audio2.mp2 -vcodec copy -acodec copy -f vob output.mpg -newaudio  Note : In order to remux a PS file with multiple audio tracks :  ffmpeg -i input.mpg -vcodec copy -acodec copy -f vob output.mpg -acodec copy -newaudio 

MPEG-2 Program Stream Demuxing

ffmpeg -i input.mpg -vcodec copy -f mpeg2video ES_Video.m2v -acodec copy -f mp2 ES_Audio.mp2  Note : This also works for files containing multiple audio tracks :  ffmpeg -i input.mpg -vcodec copy -f mpeg2video ES_Video.m2v -acodec copy -f mp2 ES_Audio1.mp2 -acodec copy -f mp2 ES_Audio2.mp2 

MPEG-2 Start Timecode

ffmpeg -i <input_file> -timecode_frame_start <start_timecode> -vcodec mpeg2video -an output.m2v  Note : Start timecode is set as number of frames. For instance, if you want to start at 18:12:36:15, you will have to set -timecode_frame_start to 1638915 ( for 25 fps content ). 

Audio Volume Modification

ffmpeg -i <input_file> -vol <audio_volume> -acodec <audio_codec> <output_file> 

Input Stream Selection

ffmpeg -i input.vob -map 0:2 -acodec aac -ab <audio_bitrate> -vn output.mp4  Transcode audio stream #0:2. 

Sub-clip Creation

ffmpeg -i <input_file> -ss <timecode> -t <timecode> -vcodec copy -acodec copy <output_file> 

Make a Video File from a Single Frame

ffmpeg -loop_input -vframes <number_of_frames> -i <input_file> <output_file> 

Simple segmentation without multi-rate or re-encode