-
-
Save tkt028/e9d662a39622cae302c503b2ef17240e to your computer and use it in GitHub Desktop.
All commands need to start with the path to your copy of ffmpeg. E.g. on windows d:\ffmpegfolder\ffmpeg.exe [command below]
Also, be sure to change input.mkv and outputFile as needed for your file.
-i input.mkv -pix_fmt yuv420p -c:v libx264 -crf 0 -g 1 -an outputFile.mp4
-i input.mkv -pix_fmt yuv420p -c:v libx264 -crf 14 -g 1 -preset faster -an outputFile.mp4
-i input.mkv -pix_fmt yuv422p10le -c:v prores_ks -profile:v 3 -vendor ap10 -an outputFile.mov
-i input.mkv -pix_fmt yuv444p10 -c:v prores_ks -profile:v 4 -vendor ap10 -an outputFile.mov
If framesize matches
-i input.mkv -pix_fmt yuv422p -c:v dnxhd -b:v 175M -an outputFile.mov
else
-i input.mkv -pix_fmt yuv422p -filter:v "scale=(sar*iw)*min(1920/(sar*iw)\,1080/ih):ih*min(1920/(sar*iw)\,1080/ih), pad=1920:1080:(1920-(sar*iw)*min(1920/(sar*iw)\,1080/ih))/2:(1080-ih*min(1920/(sar*iw)\,1080/ih))/2" -c:v dnxhd -b:v 175M -an outputFile.mov
If framesize matches
-i input.mkv -pix_fmt yuv422p -c:v dnxhd -b:v 90M -an outputFile.mov
else
-i input.mkv -pix_fmt yuv422p -filter:v "scale=(sar*iw)*min(1280/(sar*iw)\,720/ih):ih*min(1280/(sar*iw)\,720/ih), pad=1280:720:(1280-(sar*iw)*min(1280/(sar*iw)\,720/ih))/2:(720-ih*min(1280/(sar*iw)\,720/ih))/2" -c:v dnxhd -b:v 90M -an outputFile.mov
-i input.mkv -c:v utvideo -an outputFile.avi
-i input.mkv -c:v mjpeg -q:v 1 -an outputFile.avi
--ss is the start time and -t is the duration of the clip. Alternatively you can use -to to set the end timestamp of the clip.
--ss 00:01:23.123 -i input.mkv -t 20 [... remaining commands]
or
--ss 00:01:23.123 -i input.mkv -to 00:01:43.123 [... remaining commands]
-i input.mkv --ss 00:01:23.123 -t 20 [... remaining commands]
or
-i input.mkv --ss 00:01:23.123 -to 00:01:43.123 [... remaining commands]
-i input.mp4 -f segment -c:v copy -c:a copy -reset_timestamps 1 -map 0 outputFile%d.mp4
Example below splits into 5 minute chunks
-i input.mp4 -f segment -segment_time 0:05:00 -c:v copy -c:a copy -reset_timestamps 1 -map 0 outputFile%d.mp4
ffmpeg -i inputfile1.mkv [Source Settings] [Video Settings] [Audio Settings] [Container Settings] output.ext
Not usually required - ffmpeg will often choose the correct option for the codec(s) involed but some encoders can support multiple colorspaces so you sometimes need to be specific. Here are some common ones:
-pix_fmt yuv420p- same as DVD and BLu-ray footage. Common in most compressed streaming media. Use this for encoding for the web-pix_fmt yuv422p- More chroma samples, used for common intermediate and editing codecs-pix_fmt yuv444p- 'Lossless' colorspace used for high quality editing codec profiles when using raw footage.-pix_fmt yuv444p10- as above but using 10bits for the luma channel instead of 8. Better for VFX work - smoother gradients etc.
Before scaling, set the rsize algorithm: -sws_flags spline or -sws_flags lanczos will do.
Then, resize either with the scale video filter: -filter:v scale
Examples:
Scale to a particular width, auto-calculate the appropriate height:
-vf scale=1280:-1
Scale to a particular height, auto-calculate the appropriate width:
-vf scale=-1:1080
Scale to fit into a 1920x1080 frame and add borders if needed (useful for DXnHD etc):
-filter:v "scale=(sar*iw)*min(1920/(sar*iw)\,1080/ih):ih*min(1920/(sar*iw)\,1080/ih), pad=1920:1080:(1920-(sar*iw)*min(1920/(sar*iw)\,1080/ih))/2:(1080-ih*min(1920/(sar*iw)\,1080/ih))/2"
Generic pattern, for any size - just replace XXXX with the width and YYYY with the Y
-filter:v "scale=(sar*iw)*min(XXXX/(sar*iw)\,YYYY/ih):ih*min(XXXX/(sar*iw)\,YYYY/ih), pad=XXXX:YYYY:(XXXX-(sar*iw)*min(XXXX/(sar*iw)\,YYYY/ih))/2:(YYYY-ih*min(XXXX/(sar*iw)\,YYYY/ih))/2"
Choosing a video codec in ffmpeg generally uses the format -c:v followed by the codec
Here is a summary of the most important codecs for conversion:
libx264- The open-source encoder for h264 encoding (mpeg4), usually for web distribution but also can be set as a high quality lossy intermediate codec for editing on some systems.proresandprores_ks- ffmpeg's implementation of Apple's ProRes codec. Fairly industry standard for Final Cut usage - works in all pro editing systems. Main downside is the ffmpeg implementation is slow as an encoder so processing source may take longer than other options.dxnhd- Avid's editing codec. Just as well supported as ProRes and encodes faster with ffmpeg, which is a plus. However, this is profile-based so will only allow certain resolution, fps and colorspace combinations. If you source isn't exactly 720p or 1080p or what have you you will need to add borders to be able to convert using this codec.utvideo- Lossless codec, excellent quality but only really supported by Windows version of video editors. Can be sometimes slow to read making quick editing trickier but the compressed size is fairly similar to high quality lossy codecs!mjpeg- low quality traditional format that reads and writes really fast so good as a proxy source for video editors that allow low-quality proxies to be used.
The main aspects here are -crf which controlls quality (0 being lossless and 22 being the default). -g controls GOP length so setting this to 1 makes every frame a keyframe which is ideal for editing.
Lossless h264
-c:v libx264 -crf 0 -g 1
Lossy encoding with every frame a keyframe
-c:v libx264 -crf 16 -g 1
or, if you don't care about the file size that much and would rather it encode quicker:
-c:v libx264 -crf 14 -preset faster -g 1
16 is medium quality but lower numbers would be very good quality. The lower the number, the larger the file until you reach lossless.
Note the -vendor ap10 part below is only needed if working with Final Cut, but it does no harm otherwise.
High Quality 4444 - usually overkill unless doing VFX work
-pix_fmt yuv444p10 -c:v prores_ks -profile:v 4 -vendor ap10
High Quality 422 (84 Mbit/s) - use this one
-c:v prores_ks -profile:v 3 -vendor ap10
Standard Quality 422 - or this if you are low on space and need something 'good enough'
-c:v prores_ks -profile:v 2 -vendor ap10
Proxy Quality - For proxy files only, looks crappy but edits fast and is small
-c:v prores_ks -profile:v 0 -vendor ap10
First look up the bitrate you need to set on this table based on the framesize, fps and desired quality level. For each there is a standard and a high quality version. The high quality version is preferred:
| Project Format | Resolution | Frame Size | Bits | FPS | |
|---|---|---|---|---|---|
| 1080i / 59.94 | DNxHD 220 | 1920 x 1080 | 8 | 29.97 | 220M |
| 1080i / 59.94 | DNxHD 145 | 1920 x 1080 | 8 | 29.97 | 145M |
| 1080i / 50 | DNxHD 185 | 1920 x 1080 | 8 | 25 | 185M |
| 1080i / 50 | DNxHD 120 | 1920 x 1080 | 8 | 25 | 120M |
| 1080p / 25 | DNxHD 185 | 1920 x 1080 | 8 | 25 | 185M |
| 1080p / 25 | DNxHD 120 | 1920 x 1080 | 8 | 25 | 120M |
| 1080p / 25 | DNxHD 36 | 1920 x 1080 | 8 | 25 | 36M |
| 1080p / 24 | DNxHD 175 | 1920 x 1080 | 8 | 24 | 175M |
| 1080p / 24 | DNxHD 115 | 1920 x 1080 | 8 | 24 | 115M |
| 1080p / 24 | DNxHD 36 | 1920 x 1080 | 8 | 24 | 36M |
| 1080p / 23.976 | DNxHD 175 | 1920 x 1080 | 8 | 23.976 | 175M |
| 1080p / 23.976 | DNxHD 115 | 1920 x 1080 | 8 | 23.976 | 115M |
| 1080p / 23.976 | DNxHD 36 | 1920 x 1080 | 8 | 23.976 | 36M |
| 1080p / 29.7 | DNxHD 45 | 1920 x 1080 | 8 | 29.97 | 45M |
| 720p / 59.94 | DNxHD 220 | 1280x720 | 8 | 59.94 | 220M |
| 720p / 59.94 | DNxHD 145 | 1280x720 | 8 | 59.94 | 145M |
| 720p / 50 | DNxHD 175 | 1280x720 | 8 | 50 | 175M |
| 720p / 50 | DNxHD 115 | 1280x720 | 8 | 50 | 115M |
| 720p / 23.976 | DNxHD 90 | 1280x720 | 8 | 23.976 | 90M |
| 720p / 23.976 | DNxHD 60 | 1280x720 | 8 | 23.976 | 60M |
Then add the bitrate to this command (using 1080p 24fps as an example):
-c:v dnxhd -b:v 175M
UTVideo, being lossless, can generally handle any input format, so the command is simply:
-c:v utvideo
Where qscale is the quantizer to use, 1 being the highest. Still not great quality. For shitty proxy files you can drop qscale to 4 or lower to make small but ugly and fast files.
-c:v mjpeg -q:v 1
Simply add this shorthand
-an
Stereo Bitrates: 128k, 160k, 192k, 224k, 320k
224 is what iTunes uses if that gives any idea
-c:a aac -b 224k
You can determine a container format either by specifying it on the commandline with -f otherwise ffmpeg will use the filename to determine the container:
-f mp4 or simply outputFilname.mp4
-f mov or outputFilename.mov
-f mxf or outputFilename.mxf or -f mov or outputFilename.mov
-f avi or outputFilename.avi
The -crf 18 option here is very high quality. 20 would be lower quality. The default is actually 23. You can also use fractional amounts e.g. 18.5
ffmpeg -i inputfile.avi -pix_fmp yuv420p -c:v libx264 -crf 18 -preset slower -c:a aac -b:a 224k -movflags +faststart output.mp4
This is where I usually start, though for 1080p encodes this often creates a file that's way too big so I then reduce the number and find the sweet spot where it's not too large and not too bad-looking.
It's possible you may have a target device (like, I dunno, some streaming box) that chokes on high birtates you can set rate limits in addition to keeping your -crf value.
ffmpeg -i inputfile.avi -pix_fmp yuv420p -c:v libx264 -crf 18 -maxrate 7M -bufsize 10M -preset slower -c:a aac -b:a 224k -movflags +faststart output.mp4
ffmpeg [settings above except output.mp4] -metadata title="Video Title" -metadata author="Author Name" output.mp4
ffmpeg -i inputfile.avi -c:v libvpx-vp9 -crf 30 -b:v 0 -c:a libopus -b:a 224k output.webm