Differences

This shows you the differences between two versions of the page.

--- linux:multimedia:movie [2015/12/14 10:57]
niziak
+++ linux:multimedia:movie [2021/02/07 11:21] (current)
niziak
@@ Line 1: / Line 1: @@
-==== H.264 params ====
+====== Movie ======
+===== Extract frames =====
+  * Extract keyframes
+    * **-hide_banner** we are using this parameter to hide ffmpeg compilation information
+    * **-vsync vfr**: This is a parameter that tells the filter to use a variable bitrate video synchronization. If we do not use this parameter ffmpeg will fail to find only the keyframes and shoud extract other frames that can be not processed correctly.
+<code bash>ffmpeg -i VIDEO0068.mp4 -vf "select=eq(pict_type\,I)" -vsync vfr thumb%04d.bmp -hide_banner</code>
+  * Extract one frame per second
+<code bash>ffmpeg -i video.webm -vf fps=1 thumb%04d.jpg -hide_banner</code>
+  * Extract one frame per 10 seconds
+<code bash>ffmpeg -i video.webm -vf fps=1/10 thumb%04d.jpg -hide_banner</code>
+  * Extract with different quality / to bmp
+<code bash>ffmpeg -i video.webm -vf fps=1 -quality 90 thumb%04d.jpg -hide_banner</code>
+<code bash>ffmpeg -i video.webm -vf fps=1 thumb%04d.bmp -hide_banner</code>
+  * Extract only one frame
+<code bash>ffmpeg -i video.webm -ss 00:00:07.000 -vframes 1 thumb.jpg</code>
+===== Chroma subsampling =====
+[[https://en.wikipedia.org/wiki/Chroma_subsampling]]
+{{ :linux:multimedia:sampling.jpg?200|}}
+{{ :linux:multimedia:316-f6-chroma-subsampling-secondary_1.png?200|}}
+  * j:a:b - j - horizontal sampling reference (usually 4). **a** number of chroma samples (Cr, Cb) in the first row of **j** pixels. **b** number of changes of chroma samples between firs and second row of **j** pixels.
+  * 4:4:4 - (H265, H264) - no chroma subsampling - for every 4 pixels of luma, there are 4 pixels of colour
+  * 4:2:2 - 50Mbit - (H264) for every 4 pixels of luma, there are 2 pixels of colour. Keying can produce bad edges.
+  * 4:2:0 - 35MBit - (DVD, DV, JPEG, MPEG1, Blueray) - only alternatign lines are sampled for chroma. Chroma vertical resolution is halved.
+===== Stabilisation =====
+==== To remove Little shaking in 1-pass ====
+<code bash>ffmpeg -threads 8 -i $1 -vf deshake -c:a copy -c:v utvideo yuv420p $1-deshake.mkv</code>
+NOTE: No multithreading.
+==== To completely process shaky video in 2-pass| ====
+FFMpeg already contains plugins **vidstabdetect** and **vidstabtransform** [[http://public.hronopik.de/vid.stab/download.php]].
+But latest version (supports multithreading) can be downloaded as statically linked binary from: [[https://www.johnvansickle.com/ffmpeg/]]
+<code bash>
+ffmpeg -i $1 -vf vidstabdetect=shakiness=10:accuracy=15:result="tracefile.trf":show=1 -c:a copy -c:v utvideo ${PIX_FMT} ${OFILE}-vidstab1.mkv
+ffmpeg -i $1 -vf vidstabtransform=input="tracefile.trf" -c:a copy -c:v utvideo ${PIX_FMT} ${OFILE}-vidstab2.mkv
+</code>
+NOTE: output from pass1 is also written as video to see motion vectors.
+===== Lossless =====
+Use UT Video codec as better alternative to HUFYUV: [[https://en.wikipedia.org/wiki/Ut_Video_Codec_Suite]]
+  * Download for Windows: https://www.videohelp.com/software/Ut-Video-Codec-Suite
+  * Supported by ffmpeg
+<code bash>ffmpeg -i $1 -vf deshake -c:a copy -c:v utvideo -pix_fmt yuv422p $1-deshake.mkv</code>
+Comparison of file size:
+^ pixfmt ^ huffyuv ^ utvideo ^
+| 444p   |         | 4104M   |
+| 422p   | 4077M   | 3204M   |
+| 420p   |         | 2556M   |
+The same video encoded using **utvideo** @422p is 78% of **huffyuv**.
+===== MPEG2 =====
+Highest quality (qscale 4-5 is far enough)
+<code>-c:v mpeg2video -qscale:v 2</code>
+==== DVD ====
+DVD's VOBS can be simply concatenated to one big file:
+<code bash>cat ./VIDEO_TS/*.VOB | ffmpeg -i - <out_name>.<out_format></code>
+<code bash>
+#!/bin/bash -eu
+VOBS=""
+while read VOB; do
+    echo ${VOB}
+    if [ -n "${VOBS}" ]; then
+	VOBS="${VOBS}|"
+    fi
+    VOBS="${VOBS}${VOB}"
+done < <(ls -1 *.VOB)
+ffmpeg -i "concat:${VOBS}" -map 0:0 -map 0:1 -map 0:2 -c copy out.ts
+</code>
+More: [[https://newspaint.wordpress.com/2016/07/27/ripping-a-video-from-dvd-using-ffmpeg/]]
+===== H.264 params =====
 CRF (Constant Rate Factor), 0..51  (0-lossless)
-..23 very good, no blocking efect furing fast movement
+  * 17..23 very good, no blocking efect during fast movement
-- default
+  * 23 - default
+<code>-c:v libx264 -crf 25</code>
+<code>-c:v libx264 -b:v 1024k</code>
-=== Change container ===
+===== FFMpeg params =====
+.5Mbps bitrate with tolerance 300k
+<code>-b 2500k -bt 300k</code>
+==== Resize ====
+to half of size (using video filter [[https://ffmpeg.org/ffmpeg-filters.html#scale]]
+<code>-vf scale=w=iw/2:h=ih/2</code>
+==== Audio recompress ====
+channel audio (mono), audio quality 7 [[https://trac.ffmpeg.org/wiki/Encode/MP3]]
+<code>-ac 1 -c:a libmp3lame -q:a 7</code>
+===== Examples =====
+==== Change container ====
 <code>ffmpeg -i input.ts -vcodec copy -sameq -acodec copy -f matroska output.ts</code>
-==== Examples ====
+==== Copy video and compress audio ====
-=== Recompress ===
+Create mono mp3 audio stream:
+<code bash>ffmpeg -i input_file.avi -c:v copy -ac 1 -c:a libmp3lame -q:a 7 output.avi</code>
+==== Remove audio ====
+<code bash>ffmpeg -i example.mkv -c copy -an example-nosound.mkv</code>
+==== Recompress ====
+<code>
 ffmpeg -i input.mp4 -b 1000000 output.mp4
 ffmpeg -i input.mp4 -vcodec libx264 -crf 20 output.mp4
@@ Line 15: / Line 125: @@
 ffmpeg -i input.mp4 -c:v libx264 -crf 23 output.mp4
 ffmpeg -i input.ts -c:a copy -c:v libx264 -crf 18 -preset veryfast NL.mp4
+ffmpeg -i input.mkv -c:v copy -c:a
+</code>
+Limit CPU usage to 150% (2 CPUs)
+<code>cpulimit -z -e ffmpeg -l 150</code>
+==== Cut ====
+Cut input from 11:20 to 1:45:50:
+Example below (option order -i and -ss) will decode input from beginning frame by frame and starts processing at 11:20 and stops at 1:45:50.
+<code>
+ffmpeg -i input.mp4 -ss 11:20 -to 1:45:50 -c:a copy -c:v copy  output.mp4
+</code>
+Next example (option order -ss and then -i) will seek to position 11:20 using keyframes and start processing for 1:34:30.
+<code>
+ffmpeg -ss 11:20 -i input.mp4 -to 1:34:30 -c:a copy -c:v copy  output.mp4
+</code>
+This is faster solution.
+In both examples video codes is "copy" so seeking works using I-frames, so there is no speed difference. Additionally with "copy", it is possibilty
+to select time range without keyframes, so only audio will be copied and video will start from next valid key frame.
+Copy chosen streams:
+<code>ffmpeg -i test.ts -to 5:47 -map 0:0 -map0:1 -map 0:4 -c copy cut.mkv</code>
+Copy chosen streams and recompress video stream:
+<code>ffmpeg -i test.ts -to 5:47 -map 0:0 -map0:1 -map 0:4 -c copy -c:v libx264 -crf 25 cut.mkv</code>
+Copy chosen streams and resize to 50%, recompress video stream:
+<code bash>ffmpeg -i test.ts -to 5:47 -map 0:0 -map0:1 -map 0:4 -c copy -c:v libx264 -crf 25 -vf scale=w=iw/2:h=ih/2 cut.mkv</code>
+==== Deinterlace ====
+<code bash> ffmpeg -i test.ts -vf yadif -vcodec ... -acodec ... test.mp4</code>
 vidia
@@ Line 21: / Line 165: @@
 https://www.maketecheasier.com/convert-video-to-mp4-handbrake-linux/
+===== Issues =====
+ * **deprecated pixel format used, make sure you did set range correctly**
+   * when using **yuv420p** pixel format

Tools

menus and quick search

quick search

site status

Page Tools

meta data for this page

Differences