This document describes the supported formats (muxers and demuxers) provided by the libavformat library.
The libavformat library provides some generic global options, which can be set on all the muxers and demuxers. In addition each muxer or demuxer may support so-called private options, which are specific for that component.
Options may be set by specifying -option value in the
FFmpeg tools, or by setting the value explicitly in the
AVFormatContext options or using the ‘libavutil/opt.h’ API
for programmatic use.
The list of supported options follows:
Possible values:
Reduce buffering.
Set probing size in bytes, i.e. the size of the data to analyze to get stream information. A higher value will allow to detect more information in case it is dispersed into the stream, but will increase latency. Must be an integer not lesser than 32. It is 5000000 by default.
Set packet size.
Set format flags.
Possible values:
Ignore index.
Generate PTS.
Do not fill in missing values that can be exactly calculated.
Disable AVParsers, this needs +nofillin too.
Ignore DTS.
Discard corrupted frames.
Try to interleave output packets by DTS.
Do not merge side data.
Enable RTP MP4A-LATM payload.
Reduce the latency introduced by optional buffering
Allow seeking to non-keyframes on demuxer level when supported if set to 1. Default is 0.
Specify how many microseconds are analyzed to probe the input. A higher value will allow to detect more accurate information, but will increase latency. It defaults to 5,000,000 microseconds = 5 seconds.
Set decryption key.
Set max memory used for timestamp index (per stream).
Set max memory used for buffering real-time frames.
Print specific debug info.
Possible values:
Set maximum muxing or demuxing delay in microseconds.
Set number of frames used to probe fps.
Set microseconds by which audio packets should be interleaved earlier.
Set microseconds for each chunk.
Set size in bytes for each chunk.
Set error detection flags. f_err_detect is deprecated and
should be used only via the ffmpeg tool.
Possible values:
Verify embedded CRCs.
Detect bitstream specification deviations.
Detect improper bitstream length.
Abort decoding on minor error detection.
Consider things that violate the spec and have not been seen in the wild as errors.
Consider all spec non compliancies as errors.
Consider things that a sane encoder should not do as an error.
Use wallclock as timestamps.
Shift timestamps to make them non-negative. A value of 1 enables shifting, a value of 0 disables it, the default value of -1 enables shifting when required by the target format.
When shifting is enabled, all output timestamps are shifted by the same amount. Audio, video, and subtitles desynching and relative timestamp differences are preserved compared to how they would have been without shifting.
Also note that this affects only leading negative timestamps, and not non-monotonic negative timestamps.
Set number of bytes to skip before reading header and frames if set to 1. Default is 0.
Correct single timestamp overflows if set to 1. Default is 1.
Flush the underlying I/O stream after each packet. Default 1 enables it, and has the effect of reducing the latency; 0 disables it and may slightly increase performance in some cases.
Set the output time offset.
offset must be a time duration specification, see (ffmpeg-utils)time duration syntax.
The offset is added by the muxer to the output timestamps.
Specifying a positive offset means that the corresponding streams are
delayed bt the time duration specified in offset. Default value
is 0 (meaning that no offset is applied).
Format stream specifiers allow selection of one or more streams that match specific properties.
Possible forms of stream specifiers are:
Matches the stream with this index.
stream_type is one of following: ’v’ for video, ’a’ for audio, ’s’ for subtitle, ’d’ for data, and ’t’ for attachments. If stream_index is given, then it matches the stream number stream_index of this type. Otherwise, it matches all streams of this type.
If stream_index is given, then it matches the stream with number stream_index in the program with the id program_id. Otherwise, it matches all streams in the program.
Matches the stream by a format-specific ID.
The exact semantics of stream specifiers is defined by the
avformat_match_stream_specifier() function declared in the
‘libavformat/avformat.h’ header.
Demuxers are configured elements in FFmpeg that can read the multimedia streams from a particular type of file.
When you configure your FFmpeg build, all the supported demuxers
are enabled by default. You can list all available ones using the
configure option --list-demuxers.
You can disable all the demuxers using the configure option
--disable-demuxers, and selectively enable a single demuxer with
the option --enable-demuxer=DEMUXER, or disable it
with the option --disable-demuxer=DEMUXER.
The option -formats of the ff* tools will display the list of
enabled demuxers.
The description of some of the currently available demuxers follows.
Apple HTTP Live Streaming demuxer.
This demuxer presents all AVStreams from all variant streams. The id field is set to the bitrate variant index number. By setting the discard flags on AVStreams (by pressing ’a’ or ’v’ in ffplay), the caller can decide which variant streams to actually receive. The total bitrate of the variant that the stream belongs to is available in a metadata key named "variant_bitrate".
Advanced Systems Format demuxer.
This demuxer is used to demux ASF files and MMS network streams.
Do not try to resynchronize by looking for a certain optional start code.
Virtual concatenation script demuxer.
This demuxer reads a list of files and other directives from a text file and demuxes them one after the other, as if all their packet had been muxed together.
The timestamps in the files are adjusted so that the first file starts at 0 and each next file starts where the previous one finishes. Note that it is done globally and may cause gaps if all streams do not have exactly the same length.
All files must have the same streams (same codecs, same time base, etc.).
The duration of each file is used to adjust the timestamps of the next file:
if the duration is incorrect (because it was computed using the bit-rate or
because the file is truncated, for example), it can cause artifacts. The
duration directive can be used to override the duration stored in
each file.
The script is a text file in extended-ASCII, with one directive per line. Empty lines, leading spaces and lines starting with ’#’ are ignored. The following directive is recognized:
file path’Path to a file to read; special characters and spaces must be escaped with backslash or single quotes.
All subsequent directives apply to that file.
ffconcat version 1.0’Identify the script type and version. It also sets the ‘safe’ option to 1 if it was to its default -1.
To make FFmpeg recognize the format automatically, this directive must appears exactly as is (no extra space or byte-order-mark) on the very first line of the script.
duration dur’Duration of the file. This information can be specified from the file; specifying it here may be more efficient or help if the information from the file is not available or accurate.
If the duration is set for all files, then it is possible to seek in the whole concatenated video.
This demuxer accepts the following option:
If set to 1, reject unsafe file paths. A file path is considered safe if it does not contain a protocol specification and is relative and all components only contain characters from the portable character set (letters, digits, period, underscore and hyphen) and have no period at the beginning of a component.
If set to 0, any file name is accepted.
The default is -1, it is equivalent to 1 if the format was automatically probed and 0 otherwise.
Adobe Flash Video Format demuxer.
This demuxer is used to demux FLV files and RTMP network streams.
Allocate the streams according to the onMetaData array content.
The Game Music Emu library is a collection of video game music file emulators.
See http://code.google.com/p/game-music-emu/ for more information.
Some files have multiple tracks. The demuxer will pick the first track by default. The ‘track_index’ option can be used to select a different track. Track indexes start at 0. The demuxer exports the number of tracks as tracks meta data entry.
For very large files, the ‘max_size’ option may have to be adjusted.
Play media from Internet services using the quvi project.
The demuxer accepts a ‘format’ option to request a specific quality. It is by default set to best.
See http://quvi.sourceforge.net/ for more information.
FFmpeg needs to be built with --enable-libquvi for this demuxer to be
enabled.
Image file demuxer.
This demuxer reads from a list of image files specified by a pattern. The syntax and meaning of the pattern is specified by the option pattern_type.
The pattern may contain a suffix which is used to automatically determine the format of the images contained in the files.
The size, the pixel format, and the format of each image must be the same for all the files in the sequence.
This demuxer accepts the following options:
Set the frame rate for the video stream. It defaults to 25.
If set to 1, loop over the input. Default value is 0.
Select the pattern type used to interpret the provided filename.
pattern_type accepts one of the following values.
Select a sequence pattern type, used to specify a sequence of files indexed by sequential numbers.
A sequence pattern may contain the string "%d" or "%0Nd", which specifies the position of the characters representing a sequential number in each filename matched by the pattern. If the form "%d0Nd" is used, the string representing the number in each filename is 0-padded and N is the total number of 0-padded digits representing the number. The literal character ’%’ can be specified in the pattern with the string "%%".
If the sequence pattern contains "%d" or "%0Nd", the first filename of the file list specified by the pattern must contain a number inclusively contained between start_number and start_number+start_number_range-1, and all the following numbers must be sequential.
For example the pattern "img-%03d.bmp" will match a sequence of filenames of the form ‘img-001.bmp’, ‘img-002.bmp’, ..., ‘img-010.bmp’, etc.; the pattern "i%%m%%g-%d.jpg" will match a sequence of filenames of the form ‘i%m%g-1.jpg’, ‘i%m%g-2.jpg’, ..., ‘i%m%g-10.jpg’, etc.
Note that the pattern must not necessarily contain "%d" or "%0Nd", for example to convert a single image file ‘img.jpeg’ you can employ the command:
ffmpeg -i img.jpeg img.png |
Select a glob wildcard pattern type.
The pattern is interpreted like a glob() pattern. This is only
selectable if libavformat was compiled with globbing support.
Select a mixed glob wildcard/sequence pattern.
If your version of libavformat was compiled with globbing support, and
the provided pattern contains at least one glob meta character among
%*?[]{} that is preceded by an unescaped "%", the pattern is
interpreted like a glob() pattern, otherwise it is interpreted
like a sequence pattern.
All glob special characters %*?[]{} must be prefixed
with "%". To escape a literal "%" you shall use "%%".
For example the pattern foo-%*.jpeg will match all the
filenames prefixed by "foo-" and terminating with ".jpeg", and
foo-%?%?%?.jpeg will match all the filenames prefixed with
"foo-", followed by a sequence of three characters, and terminating
with ".jpeg".
This pattern type is deprecated in favor of glob and sequence.
Default value is glob_sequence.
Set the pixel format of the images to read. If not specified the pixel format is guessed from the first image file in the sequence.
Set the index of the file matched by the image file pattern to start to read from. Default value is 0.
Set the index interval range to check when looking for the first image file in the sequence, starting from start_number. Default value is 5.
If set to 1, will set frame timestamp to modification time of image file. Note that monotonity of timestamps is not provided: images go in the same order as without this option. Default value is 0.
Set the video size of the images to read. If not specified the video size is guessed from the first image file in the sequence.
ffmpeg for creating a video from the images in the file
sequence ‘img-001.jpeg’, ‘img-002.jpeg’, ..., assuming an
input frame rate of 10 frames per second:
ffmpeg -framerate 10 -i 'img-%03d.jpeg' out.mkv |
ffmpeg -framerate 10 -start_number 100 -i 'img-%03d.jpeg' out.mkv |
ffmpeg -framerate 10 -pattern_type glob -i "*.png" out.mkv |
MPEG-2 transport stream demuxer.
Overrides teletext packet PTS and DTS values with the timestamps calculated from the PCR of the first program which the teletext stream is part of and is not discarded. Default value is 1, set this option to 0 if you want your teletext packet PTS and DTS values untouched.
Raw video demuxer.
This demuxer allows to read raw video data. Since there is no header specifying the assumed video parameters, the user must specify them in order to be able to decode the data correctly.
This demuxer accepts the following options:
Set input video frame rate. Default value is 25.
Set the input video pixel format. Default value is yuv420p.
Set the input video size. This value must be specified explicitly.
For example to read a rawvideo file ‘input.raw’ with
ffplay, assuming a pixel format of rgb24, a video
size of 320x240, and a frame rate of 10 images per second, use
the command:
ffplay -f rawvideo -pixel_format rgb24 -video_size 320x240 -framerate 10 input.raw |
SBaGen script demuxer.
This demuxer reads the script language used by SBaGen http://uazu.net/sbagen/ to generate binaural beats sessions. A SBG script looks like that:
-SE a: 300-2.5/3 440+4.5/0 b: 300-2.5/0 440+4.5/3 off: - NOW == a +0:07:00 == b +0:14:00 == a +0:21:00 == b +0:30:00 off |
A SBG script can mix absolute and relative timestamps. If the script uses either only absolute timestamps (including the script start time) or only relative ones, then its layout is fixed, and the conversion is straightforward. On the other hand, if the script mixes both kind of timestamps, then the NOW reference for relative timestamps will be taken from the current time of day at the time the script is read, and the script layout will be frozen according to that reference. That means that if the script is directly played, the actual times will match the absolute timestamps up to the sound controller’s clock accuracy, but if the user somehow pauses the playback or seeks, all times will be shifted accordingly.
JSON captions used for TED Talks.
TED does not provide links to the captions, but they can be guessed from the page. The file ‘tools/bookmarklets.html’ from the FFmpeg source tree contains a bookmarklet to expose them.
This demuxer accepts the following option:
Set the start time of the TED talk, in milliseconds. The default is 15000 (15s). It is used to sync the captions with the downloadable videos, because they include a 15s intro.
Example: convert the captions to a format most players understand:
ffmpeg -i http://www.ted.com/talks/subtitles/id/1/lang/en talk1-en.srt |
Muxers are configured elements in FFmpeg which allow writing multimedia streams to a particular type of file.
When you configure your FFmpeg build, all the supported muxers
are enabled by default. You can list all available muxers using the
configure option --list-muxers.
You can disable all the muxers with the configure option
--disable-muxers and selectively enable / disable single muxers
with the options --enable-muxer=MUXER /
--disable-muxer=MUXER.
The option -formats of the ff* tools will display the list of
enabled muxers.
A description of some of the currently available muxers follows.
Audio Interchange File Format muxer.
It accepts the following options:
Enable ID3v2 tags writing when set to 1. Default is 0 (disabled).
Select ID3v2 version to write. Currently only version 3 and 4 (aka. ID3v2.3 and ID3v2.4) are supported. The default is version 4.
CRC (Cyclic Redundancy Check) testing format.
This muxer computes and prints the Adler-32 CRC of all the input audio and video frames. By default audio frames are converted to signed 16-bit raw audio and video frames to raw video before computing the CRC.
The output of the muxer consists of a single line of the form: CRC=0xCRC, where CRC is a hexadecimal number 0-padded to 8 digits containing the CRC for all the decoded input frames.
See also the framecrc muxer.
For example to compute the CRC of the input, and store it in the file ‘out.crc’:
ffmpeg -i INPUT -f crc out.crc |
You can print the CRC to stdout with the command:
ffmpeg -i INPUT -f crc - |
You can select the output format of each frame with ffmpeg by
specifying the audio and video codec and format. For example to
compute the CRC of the input audio converted to PCM unsigned 8-bit
and the input video converted to MPEG-2 video, use the command:
ffmpeg -i INPUT -c:a pcm_u8 -c:v mpeg2video -f crc - |
Per-packet CRC (Cyclic Redundancy Check) testing format.
This muxer computes and prints the Adler-32 CRC for each audio and video packet. By default audio frames are converted to signed 16-bit raw audio and video frames to raw video before computing the CRC.
The output of the muxer consists of a line for each audio and video packet of the form:
stream_index, packet_dts, packet_pts, packet_duration, packet_size, 0xCRC |
CRC is a hexadecimal number 0-padded to 8 digits containing the CRC of the packet.
For example to compute the CRC of the audio and video frames in ‘INPUT’, converted to raw audio and video packets, and store it in the file ‘out.crc’:
ffmpeg -i INPUT -f framecrc out.crc |
To print the information to stdout, use the command:
ffmpeg -i INPUT -f framecrc - |
With ffmpeg, you can select the output format to which the
audio and video frames are encoded before computing the CRC for each
packet by specifying the audio and video codec. For example, to
compute the CRC of each decoded input audio frame converted to PCM
unsigned 8-bit and of each decoded input video frame converted to
MPEG-2 video, use the command:
ffmpeg -i INPUT -c:a pcm_u8 -c:v mpeg2video -f framecrc - |
See also the crc muxer.
Per-packet MD5 testing format.
This muxer computes and prints the MD5 hash for each audio and video packet. By default audio frames are converted to signed 16-bit raw audio and video frames to raw video before computing the hash.
The output of the muxer consists of a line for each audio and video packet of the form:
stream_index, packet_dts, packet_pts, packet_duration, packet_size, MD5 |
MD5 is a hexadecimal number representing the computed MD5 hash for the packet.
For example to compute the MD5 of the audio and video frames in ‘INPUT’, converted to raw audio and video packets, and store it in the file ‘out.md5’:
ffmpeg -i INPUT -f framemd5 out.md5 |
To print the information to stdout, use the command:
ffmpeg -i INPUT -f framemd5 - |
See also the md5 muxer.
Animated GIF muxer.
It accepts the following options:
Set the number of times to loop the output. Use -1 for no loop, 0
for looping indefinitely (default).
Force the delay (expressed in centiseconds) after the last frame. Each frame
ends with a delay until the next frame. The default is -1, which is a
special value to tell the muxer to re-use the previous delay. In case of a
loop, you might want to customize this value to mark a pause for instance.
For example, to encode a gif looping 10 times, with a 5 seconds delay between the loops:
ffmpeg -i INPUT -loop 10 -final_delay 500 out.gif |
Note 1: if you wish to extract the frames in separate GIF files, you need to force the image2 muxer:
ffmpeg -i INPUT -c:v gif -f image2 "out%d.gif" |
Note 2: the GIF format has a very small time base: the delay between two frames can not be smaller than one centi second.
Apple HTTP Live Streaming muxer that segments MPEG-TS according to the HTTP Live Streaming (HLS) specification.
It creates a playlist file and numbered segment files. The output filename specifies the playlist filename; the segment filenames receive the same basename as the playlist, a sequential number and a .ts extension.
For example, to convert an input file with ffmpeg:
ffmpeg -i in.nut out.m3u8 |
See also the segment muxer, which provides a more generic and flexible implementation of a segmenter, and can be used to perform HLS segmentation.
This muxer supports the following options:
Set the segment length in seconds. Default value is 2.
Set the maximum number of playlist entries. If set to 0 the list file will contain all the segments. Default value is 5.
Set the number after which the segment filename number (the number specified in each segment file) wraps. If set to 0 the number will be never wrapped. Default value is 0.
This option is useful to avoid to fill the disk with many segment files, and limits the maximum number of segment files written to disk to wrap.
Start the playlist sequence number from number. Default value is 0.
Note that the playlist sequence number must be unique for each segment and it is not to be confused with the segment filename sequence number which can be cyclic, for example if the ‘wrap’ option is specified.
ICO file muxer.
Microsoft’s icon file format (ICO) has some strict limitations that should be noted:
BMP Bit Depth FFmpeg Pixel Format 1bit pal8 4bit pal8 8bit pal8 16bit rgb555le 24bit bgr24 32bit bgra |
Image file muxer.
The image file muxer writes video frames to image files.
The output filenames are specified by a pattern, which can be used to produce sequentially numbered series of files. The pattern may contain the string "%d" or "%0Nd", this string specifies the position of the characters representing a numbering in the filenames. If the form "%0Nd" is used, the string representing the number in each filename is 0-padded to N digits. The literal character ’%’ can be specified in the pattern with the string "%%".
If the pattern contains "%d" or "%0Nd", the first filename of the file list specified will contain the number 1, all the following numbers will be sequential.
The pattern may contain a suffix which is used to automatically determine the format of the image files to write.
For example the pattern "img-%03d.bmp" will specify a sequence of filenames of the form ‘img-001.bmp’, ‘img-002.bmp’, ..., ‘img-010.bmp’, etc. The pattern "img%%-%d.jpg" will specify a sequence of filenames of the form ‘img%-1.jpg’, ‘img%-2.jpg’, ..., ‘img%-10.jpg’, etc.
The following example shows how to use ffmpeg for creating a
sequence of files ‘img-001.jpeg’, ‘img-002.jpeg’, ...,
taking one image every second from the input video:
ffmpeg -i in.avi -vsync 1 -r 1 -f image2 'img-%03d.jpeg' |
Note that with ffmpeg, if the format is not specified with the
-f option and the output filename specifies an image file
format, the image2 muxer is automatically selected, so the previous
command can be written as:
ffmpeg -i in.avi -vsync 1 -r 1 'img-%03d.jpeg' |
Note also that the pattern must not necessarily contain "%d" or "%0Nd", for example to create a single image file ‘img.jpeg’ from the input video you can employ the command:
ffmpeg -i in.avi -f image2 -frames:v 1 img.jpeg |
The ‘strftime’ option allows you to expand the filename with
date and time information. Check the documentation of
the strftime() function for the syntax.
For example to generate image files from the strftime()
"%Y-%m-%d_%H-%M-%S" pattern, the following ffmpeg command
can be used:
ffmpeg -f v4l2 -r 1 -i /dev/video0 -f image2 -strftime 1 "%Y-%m-%d_%H-%M-%S.jpg" |
Start the sequence from the specified number. Default value is 1. Must be a non-negative number.
If set to 1, the filename will always be interpreted as just a filename, not a pattern, and the corresponding file will be continuously overwritten with new images. Default value is 0.
If set to 1, expand the filename with date and time information from
strftime(). Default value is 0.
The image muxer supports the .Y.U.V image file format. This format is special in that that each image frame consists of three files, for each of the YUV420P components. To read or write this image file format, specify the name of the ’.Y’ file. The muxer will automatically open the ’.U’ and ’.V’ files as required.
Matroska container muxer.
This muxer implements the matroska and webm container specs.
The recognized metadata settings in this muxer are:
Set title name provided to a single track.
Specify the language of the track in the Matroska languages form.
The language can be either the 3 letters bibliographic ISO-639-2 (ISO 639-2/B) form (like "fre" for French), or a language code mixed with a country code for specialities in languages (like "fre-ca" for Canadian French).
Set stereo 3D video layout of two views in a single video track.
The following values are recognized:
video is not stereo
Both views are arranged side by side, Left-eye view is on the left
Both views are arranged in top-bottom orientation, Left-eye view is at bottom
Both views are arranged in top-bottom orientation, Left-eye view is on top
Each view is arranged in a checkerboard interleaved pattern, Left-eye view being first
Each view is arranged in a checkerboard interleaved pattern, Right-eye view being first
Each view is constituted by a row based interleaving, Right-eye view is first row
Each view is constituted by a row based interleaving, Left-eye view is first row
Both views are arranged in a column based interleaving manner, Right-eye view is first column
Both views are arranged in a column based interleaving manner, Left-eye view is first column
All frames are in anaglyph format viewable through red-cyan filters
Both views are arranged side by side, Right-eye view is on the left
All frames are in anaglyph format viewable through green-magenta filters
Both eyes laced in one Block, Left-eye view is first
Both eyes laced in one Block, Right-eye view is first
For example a 3D WebM clip can be created using the following command line:
ffmpeg -i sample_left_right_clip.mpg -an -c:v libvpx -metadata stereo_mode=left_right -y stereo_clip.webm |
This muxer supports the following options:
By default, this muxer writes the index for seeking (called cues in Matroska terms) at the end of the file, because it cannot know in advance how much space to leave for the index at the beginning of the file. However for some use cases – e.g. streaming where seeking is possible but slow – it is useful to put the index at the beginning of the file.
If this option is set to a non-zero value, the muxer will reserve a given amount of space in the file header and then try to write the cues there when the muxing finishes. If the available space does not suffice, muxing will fail. A safe size for most use cases should be about 50kB per hour of video.
Note that cues are only written if the output is seekable and this option will have no effect if it is not.
MD5 testing format.
This muxer computes and prints the MD5 hash of all the input audio and video frames. By default audio frames are converted to signed 16-bit raw audio and video frames to raw video before computing the hash.
The output of the muxer consists of a single line of the form: MD5=MD5, where MD5 is a hexadecimal number representing the computed MD5 hash.
For example to compute the MD5 hash of the input converted to raw audio and video, and store it in the file ‘out.md5’:
ffmpeg -i INPUT -f md5 out.md5 |
You can print the MD5 to stdout with the command:
ffmpeg -i INPUT -f md5 - |
See also the framemd5 muxer.
MOV/MP4/ISMV (Smooth Streaming) muxer.
The mov/mp4/ismv muxer supports fragmentation. Normally, a MOV/MP4
file has all the metadata about all packets stored in one location
(written at the end of the file, it can be moved to the start for
better playback by adding faststart to the movflags, or
using the qt-faststart tool). A fragmented
file consists of a number of fragments, where packets and metadata
about these packets are stored together. Writing a fragmented
file has the advantage that the file is decodable even if the
writing is interrupted (while a normal MOV/MP4 is undecodable if
it is not properly finished), and it requires less memory when writing
very long files (since writing normal MOV/MP4 files stores info about
every single packet in memory until the file is closed). The downside
is that it is less compatible with other applications.
Fragmentation is enabled by setting one of the AVOptions that define how to cut the file into fragments:
Reserves space for the moov atom at the beginning of the file instead of placing the moov atom at the end. If the space reserved is insufficient, muxing will fail.
Start a new fragment at each video keyframe.
Create fragments that are duration microseconds long.
Create fragments that contain up to size bytes of payload data.
Allow the caller to manually choose when to cut fragments, by
calling av_write_frame(ctx, NULL) to write a fragment with
the packets written so far. (This is only useful with other
applications integrating libavformat, not from ffmpeg.)
Don’t create fragments that are shorter than duration microseconds long.
If more than one condition is specified, fragments are cut when
one of the specified conditions is fulfilled. The exception to this is
-min_frag_duration, which has to be fulfilled for any of the other
conditions to apply.
Additionally, the way the output file is written can be adjusted through a few other options:
Write an initial moov atom directly at the start of the file, without describing any samples in it. Generally, an mdat/moov pair is written at the start of the file, as a normal MOV/MP4 file, containing only a short portion of the file. With this option set, there is no initial mdat atom, and the moov atom only describes the tracks but has a zero duration.
Files written with this option set do not work in QuickTime. This option is implicitly set when writing ismv (Smooth Streaming) files.
Write a separate moof (movie fragment) atom for each track. Normally, packets for all tracks are written in a moof atom (which is slightly more efficient), but with this option set, the muxer writes one moof/mdat pair for each track, making it easier to separate tracks.
This option is implicitly set when writing ismv (Smooth Streaming) files.
Run a second pass moving the index (moov atom) to the beginning of the file. This operation can take a while, and will not work in various situations such as fragmented output, thus it is not enabled by default.
Add RTP hinting tracks to the output file.
Smooth Streaming content can be pushed in real time to a publishing point on IIS with this muxer. Example:
ffmpeg -re <normal input/transcoding options> -movflags isml+frag_keyframe -f ismv http://server/publishingpoint.isml/Streams(Encoder1) |
The MP3 muxer writes a raw MP3 stream with an ID3v2 header at the beginning and
optionally an ID3v1 tag at the end. ID3v2.3 and ID3v2.4 are supported, the
id3v2_version option controls which one is used. Setting
id3v2_version to 0 will disable the ID3v2 header completely. The legacy
ID3v1 tag is not written by default, but may be enabled with the
write_id3v1 option.
The muxer may also write a Xing frame at the beginning, which contains the
number of frames in the file. It is useful for computing duration of VBR files.
The Xing frame is written if the output stream is seekable and if the
write_xing option is set to 1 (the default).
The muxer supports writing ID3v2 attached pictures (APIC frames). The pictures are supplied to the muxer in form of a video stream with a single packet. There can be any number of those streams, each will correspond to a single APIC frame. The stream metadata tags title and comment map to APIC description and picture type respectively. See http://id3.org/id3v2.4.0-frames for allowed picture types.
Note that the APIC frames must be written at the beginning, so the muxer will buffer the audio frames until it gets all the pictures. It is therefore advised to provide the pictures as soon as possible to avoid excessive buffering.
Examples:
Write an mp3 with an ID3v2.3 header and an ID3v1 footer:
ffmpeg -i INPUT -id3v2_version 3 -write_id3v1 1 out.mp3 |
To attach a picture to an mp3 file select both the audio and the picture stream
with map:
ffmpeg -i input.mp3 -i cover.png -c copy -map 0 -map 1 -metadata:s:v title="Album cover" -metadata:s:v comment="Cover (Front)" out.mp3 |
Write a "clean" MP3 without any extra features:
ffmpeg -i input.wav -write_xing 0 -id3v2_version 0 out.mp3 |
MPEG transport stream muxer.
This muxer implements ISO 13818-1 and part of ETSI EN 300 468.
The recognized metadata settings in mpegts muxer are service_provider
and service_name. If they are not set the default for
service_provider is "FFmpeg" and the default for
service_name is "Service01".
The muxer options are:
Set the original_network_id (default 0x0001). This is unique identifier of a network in DVB. Its main use is in the unique identification of a service through the path Original_Network_ID, Transport_Stream_ID.
Set the transport_stream_id (default 0x0001). This identifies a transponder in DVB.
Set the service_id (default 0x0001) also known as program in DVB.
Set the first PID for PMT (default 0x1000, max 0x1f00).
Set the first PID for data packets (default 0x0100, max 0x0f00).
Enable m2ts mode if set to 1. Default value is -1 which disables m2ts mode.
Set muxrate.
Set minimum PES packet payload in bytes.
Set flags (see below).
Preserve original timestamps, if value is set to 1. Default value is -1, which results in shifting timestamps so that they start from 0.
Set PAT, PMT and SDT version (default 0, valid values are from 0 to 31, inclusively). This option allows updating stream structure so that standard consumer may detect the change. To do so, reopen output AVFormatContext (in case of API usage) or restart ffmpeg instance, cyclically changing tables_version value:
ffmpeg -i source1.ts -codec copy -f mpegts -tables_version 0 udp://1.1.1.1:1111 ffmpeg -i source2.ts -codec copy -f mpegts -tables_version 1 udp://1.1.1.1:1111 ... ffmpeg -i source3.ts -codec copy -f mpegts -tables_version 31 udp://1.1.1.1:1111 ffmpeg -i source1.ts -codec copy -f mpegts -tables_version 0 udp://1.1.1.1:1111 ffmpeg -i source2.ts -codec copy -f mpegts -tables_version 1 udp://1.1.1.1:1111 ... |
Option mpegts_flags may take a set of such flags:
Reemit PAT/PMT before writing the next packet.
Use LATM packetization for AAC.
ffmpeg -i file.mpg -c copy \
-mpegts_original_network_id 0x1122 \
-mpegts_transport_stream_id 0x3344 \
-mpegts_service_id 0x5566 \
-mpegts_pmt_start_pid 0x1500 \
-mpegts_start_pid 0x150 \
-metadata service_provider="Some provider" \
-metadata service_name="Some Channel" \
-y out.ts
|
Null muxer.
This muxer does not generate any output file, it is mainly useful for testing or benchmarking purposes.
For example to benchmark decoding with ffmpeg you can use the
command:
ffmpeg -benchmark -i INPUT -f null out.null |
Note that the above command does not read or write the ‘out.null’
file, but specifying the output file is required by the ffmpeg
syntax.
Alternatively you can write the command as:
ffmpeg -benchmark -i INPUT -f null - |
Ogg container muxer.
Preferred page duration, in microseconds. The muxer will attempt to create pages that are approximately duration microseconds long. This allows the user to compromise between seek granularity and container overhead. The default is 1 second. A value of 0 will fill all segments, making pages as large as possible. A value of 1 will effectively use 1 packet-per-page in most situations, giving a small seek granularity at the cost of additional container overhead.
Basic stream segmenter.
This muxer outputs streams to a number of separate files of nearly fixed duration. Output filename pattern can be set in a fashion similar to image2.
stream_segment is a variant of the muxer used to write to
streaming output formats, i.e. which do not require global headers,
and is recommended for outputting e.g. to MPEG transport stream segments.
ssegment is a shorter alias for stream_segment.
Every segment starts with a keyframe of the selected reference stream, which is set through the ‘reference_stream’ option.
Note that if you want accurate splitting for a video file, you need to make the input key frames correspond to the exact splitting times expected by the segmenter, or the segment muxer will start the new segment with the key frame found next after the specified start time.
The segment muxer works best with a single constant frame rate video.
Optionally it can generate a list of the created segments, by setting the option segment_list. The list type is specified by the segment_list_type option. The entry filenames in the segment list are set by default to the basename of the corresponding segment files.
See also the hls muxer, which provides a more specific implementation for HLS segmentation.
The segment muxer supports the following options:
Set the reference stream, as specified by the string specifier.
If specifier is set to auto, the reference is choosen
automatically. Otherwise it must be a stream specifier (see the “Stream
specifiers” chapter in the ffmpeg manual) which specifies the
reference stream. The default value is auto.
Override the inner container format, by default it is guessed by the filename extension.
Generate also a listfile named name. If not specified no listfile is generated.
Set flags affecting the segment list generation.
It currently supports the following flags:
Allow caching (only affects M3U8 list files).
Allow live-friendly file generation.
Update the list file so that it contains at most the last size segments. If 0 the list file will contain all the segments. Default value is 0.
Set prefix to prepend to the name of each entry filename. By default no prefix is applied.
Specify the format for the segment list file.
The following values are recognized:
Generate a flat list for the created segments, one segment per line.
Generate a list for the created segments, one segment per line, each line matching the format (comma-separated values):
segment_filename,segment_start_time,segment_end_time |
segment_filename is the name of the output file generated by the muxer according to the provided pattern. CSV escaping (according to RFC4180) is applied if required.
segment_start_time and segment_end_time specify the segment start and end time expressed in seconds.
A list file with the suffix ".csv" or ".ext" will
auto-select this format.
‘ext’ is deprecated in favor or ‘csv’.
Generate an ffconcat file for the created segments. The resulting file can be read using the FFmpeg concat demuxer.
A list file with the suffix ".ffcat" or ".ffconcat" will
auto-select this format.
Generate an extended M3U8 file, version 3, compliant with http://tools.ietf.org/id/draft-pantos-http-live-streaming.
A list file with the suffix ".m3u8" will auto-select this format.
If not specified the type is guessed from the list file name suffix.
Set segment duration to time, the value must be a duration specification. Default value is "2". See also the ‘segment_times’ option.
Note that splitting may not be accurate, unless you force the reference stream key-frames at the given time. See the introductory notice and the examples below.
Specify the accuracy time when selecting the start time for a segment, expressed as a duration specification. Default value is "0".
When delta is specified a key-frame will start a new segment if its PTS satisfies the relation:
PTS >= start_time - time_delta |
This option is useful when splitting video content, which is always split at GOP boundaries, in case a key frame is found just before the specified split time.
In particular may be used in combination with the ‘ffmpeg’ option force_key_frames. The key frame times specified by force_key_frames may not be set accurately because of rounding issues, with the consequence that a key frame time may result set just before the specified time. For constant frame rate videos a value of 1/(2*frame_rate) should address the worst case mismatch between the specified time and the time set by force_key_frames.
Specify a list of split points. times contains a list of comma separated duration specifications, in increasing order. See also the ‘segment_time’ option.
Specify a list of split video frame numbers. frames contains a list of comma separated integer numbers, in increasing order.
This option specifies to start a new segment whenever a reference stream key frame is found and the sequential number (starting from 0) of the frame is greater or equal to the next value in the list.
Wrap around segment index once it reaches limit.
Set the sequence number of the first segment. Defaults to 0.
Reset timestamps at the begin of each segment, so that each segment
will start with near-zero timestamps. It is meant to ease the playback
of the generated segments. May not work with some combinations of
muxers/codecs. It is set to 0 by default.
Specify timestamp offset to apply to the output packet timestamps. The argument must be a time duration specification, and defaults to 0.
ffmpeg -i in.mkv -codec copy -map 0 -f segment -segment_list out.list out%03d.nut |
ffmpeg -i in.mkv -codec copy -map 0 -f segment -segment_list out.csv -segment_times 1,2,3,5,8,13,21 out%03d.nut |
ffmpeg ‘force_key_frames’
option to force key frames in the input at the specified location, together
with the segment option ‘segment_time_delta’ to account for
possible roundings operated when setting key frame times.
ffmpeg -i in.mkv -force_key_frames 1,2,3,5,8,13,21 -codec:v mpeg4 -codec:a pcm_s16le -map 0 \ -f segment -segment_list out.csv -segment_times 1,2,3,5,8,13,21 -segment_time_delta 0.05 out%03d.nut |
In order to force key frames on the input file, transcoding is required.
ffmpeg -i in.mkv -codec copy -map 0 -f segment -segment_list out.csv -segment_frames 100,200,300,500,800 out%03d.nut |
libx264
and libfaac encoders:
ffmpeg -i in.mkv -map 0 -codec:v libx264 -codec:a libfaac -f ssegment -segment_list out.list out%03d.ts |
ffmpeg -re -i in.mkv -codec copy -map 0 -f segment -segment_list playlist.m3u8 \ -segment_list_flags +live -segment_time 10 out%03d.mkv |
The tee muxer can be used to write the same data to several files or any other kind of muxer. It can be used, for example, to both stream a video to the network and save it to disk at the same time.
It is different from specifying several outputs to the ffmpeg
command-line tool because the audio and video data will be encoded only once
with the tee muxer; encoding can be a very expensive process. It is not
useful when using the libavformat API directly because it is then possible
to feed the same packets to several muxers directly.
The slave outputs are specified in the file name given to the muxer, separated by ’|’. If any of the slave name contains the ’|’ separator, leading or trailing spaces or any special character, it must be escaped (see (ffmpeg-utils)quoting_and_escaping).
Muxer options can be specified for each slave by prepending them as a list of key=value pairs separated by ’:’, between square brackets. If the options values contain a special character or the ’:’ separator, they must be escaped; note that this is a second level escaping.
The following special options are also recognized:
Specify the format name. Useful if it cannot be guessed from the output name suffix.
Specify a list of bitstream filters to apply to the specified output.
It is possible to specify to which streams a given bitstream filter
applies, by appending a stream specifier to the option separated by
/. spec must be a stream specifier (see Format stream specifiers). If the stream specifier is not specified, the
bistream filters will be applied to all streams in the output.
Several bitstream filters can be specified, separated by ",".
Select the streams that should be mapped to the slave output, specified by a stream specifier. If not specified, this defaults to all the input streams.
ffmpeg -i ... -c:v libx264 -c:a mp2 -f tee -map 0:v -map 0:a "archive-20121107.mkv|[f=mpegts]udp://10.0.1.255:1234/" |
ffmpeg to encode the input, and send the output
to three different destinations. The dump_extra bitstream
filter is used to add extradata information to all the output video
keyframes packets, as requested by the MPEG-TS format. The select
option is applied to ‘out.aac’ in order to make it contain only
audio packets.
ffmpeg -i ... -map 0 -flags +global_header -c:v libx264 -c:a aac -strict experimental
-f tee "[bsfs/v=dump_extra]out.ts|[movflags=+faststart]out.mp4|[select=a]out.aac"
|
a:1 for the audio output. Note
that a second level escaping must be performed, as ":" is a special
character used to separate options.
ffmpeg -i ... -map 0 -flags +global_header -c:v libx264 -c:a aac -strict experimental
-f tee "[bsfs/v=dump_extra]out.ts|[movflags=+faststart]out.mp4|[select=\'a:1\']out.aac"
|
Note: some codecs may need different options depending on the output format; the auto-detection of this can not work with the tee muxer. The main example is the ‘global_header’ flag.
FFmpeg is able to dump metadata from media files into a simple UTF-8-encoded INI-like text file and then load it back using the metadata muxer/demuxer.
The file format is as follows:
A ffmetadata file might look like this:
;FFMETADATA1 title=bike\\shed ;this is a comment artist=FFmpeg troll team [CHAPTER] TIMEBASE=1/1000 START=0 #chapter ends at 0:01:00 END=60000 title=chapter \#1 [STREAM] title=multi\ line |
By using the ffmetadata muxer and demuxer it is possible to extract metadata from an input file to an ffmetadata file, and then transcode the file into an output file with the edited ffmetadata file.
Extracting an ffmetadata file with ‘ffmpeg’ goes as follows:
ffmpeg -i INPUT -f ffmetadata FFMETADATAFILE |
Reinserting edited metadata information from the FFMETADATAFILE file can be done as:
ffmpeg -i INPUT -i FFMETADATAFILE -map_metadata 1 -codec copy OUTPUT |
ffmpeg, ffplay, ffprobe, ffserver, libavformat
The FFmpeg developers.
For details about the authorship, see the Git history of the project
(git://source.ffmpeg.org/ffmpeg), e.g. by typing the command
git log in the FFmpeg source directory, or browsing the
online repository at http://source.ffmpeg.org.
Maintainers for the specific components are listed in the file ‘MAINTAINERS’ in the source code tree.