All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

• plugins.rawdata_pillow - added pillow plugin to convert video/image stream to/from Pillow Image object

• caps to cache the outputs of its functions

• FFmpeg v8 and v8.1 compatibility
• FFmpeg version may start with 'n' to support a version name with a release branch name
• caps subpackage - fixed several bugs
• plugins.finder_ffdl - fixed logic to check existence of ffmpeg binaries

• audio.filter() and image.filter(): fixed to process their sp_kwargs argument
• numPy plugin: fixed to convert the array view to bytes (fixes stdin operations in Linux)
• probe: fixed to accept any stringifiable object as url also cleaned up the url processing

• allow writers' extra_inputs arguments to be str or tuple[str, dict|None]
• probe functions accepts PathLike object as the media url

• configure.add_urls() to handle extra_inputs argument processing

• probe._exec() to decode the error message sent from ffprobe
• filtergraph.Filter.add_labels(): fixed a syntax error

• caps submodule - update format regex for v7 compatibility
• filter functions accept expr=None for implicit filters (e.g., reshape and resample)
• probe.video/audio_stream_basic(): added keep_str_values bool arg
• probe._items_to_numeric(): not convert hex values
• probe: added custom audio & video info probe functions - added _audio_info() and _video_info()
• audio.read() & SimpleAudioReader: switched to use _audio_info()
• image.read() video.read(), & SimpleVideoReader: switched to use _video_info()
• probe: url argument to support bytes-like object
• probe._items_to_numeric(): added to convert hex int and a:b ratio fraction
• probe:_add_select_streams: convert stream spec to str (if int)
• probe: added keep_str_values argument to ffprobe calling functions
• probe: added keep_optional_fields argument to ffprobe calling functions
• probe: added sp_kwargs argument to all ffprobe running functions
• probe: added frames() to __all__
• open(): major changes in input arguments
• Filter classes: Renamed constructor arguments
• LoggerThread: pass ffmpeg logs to debug log
• open(): no longer a context manager
• Popen.send_signal: defaults to perform ctrl-c
• SimpleReaderBase.close: poll before calling terminate

• AviMediaReader: fixed Issue #46 - a bug in AviReaderThread.wait
• AviMediaReader: fixed __next__ behavior (now throws StopIteration)
• AviReaderThread: fixed buffer concatenation bug
• compose_filter_args: fixed escaping commas
• probe._resolve_entries(): fixed return value
• probe.query: fixed stream request id
• probe: fixedcache_output=True operation
• probe.query(): fixed a logic to return a single stream
• probe:_add_show_entries: fixed error if entries is bool
• Popen: fixed progmon's cancel operation (ctrl-c)
• ProgressMonitorThread: call cancelfun() only once

• probe: removed auto-caching ffprobe output

• util.parser.compose: fixed a bug composing FFmpeg arguments with an option with different values to two or more stream
• streams.SimpleStreams.close: prevent OSError even if pipes to ffmpeg process fail to close
• threading.LoggerThread.run: fixed to running logger thread when stderr not available
• open to convert unused rate argument to ffmpeg option

• No longer uses root logger. Creates and uses "ffmpegio" logger
• Tweaked to be compatible with FFmpeg v6.1
• Updated handling of repeated option values (overall + per-stream) so that per-stream value gets preference

• probe.query: return only the requested fields when retrieving from a cached result
• filtergraph: can be imported even if ffmpeg executable is not present

• path.ffmpeg & path.ffprobe: throw FFmpegNotFound if the executable not found

• probe.query: return a list of info of multiple streams if stream index is not specified
• probe.query: sped up querying specific fields (avoids full query)
• probe.query: no longer errors if a field is invalid

• transcode() can output to stdout

• Fixed show_log handling in many functions

• Dropped support for Python 3.7 but added 3.12
• Dropped setup.py. All setup metadata in pyproject.toml

• SimpleStream writer/filter removed byte-casting of input buffer

• plugins.rawdata_bytes casting 'buffers' as memoryview has erroneous effect; now assume to receive bytes/bytearray

• default plugins to return None if not compatible
• analyze.BlurDetect.log check name not key

• filtergraph submodule for building complex filtergraphs
• utils.parser.compose() - support dict value for -metadata output option
• sp_kwargs argument to all the read, write, and filter functions and classes.

• SimpleStreams.SimpleReaderBase.readinto() - fixed plugin invocation bug

• added missing show_log argument to image.filter()

• filtergraph module with Filter, Chain, and Graph classes as FFmpeg filtergraph construction and manipulation tools

• ffmpeg_ver for easy access to linked FFmpeg version

• FFmpegioError exception class

• utils.is_stream_spec: new option file_index=None to search with or without filter index

• consolidate representation of input/output/internal links for filtergraphs, affecting util.filter.parse_graph() & util.filter.compose_graph()
• corrected function names util.*spec_stream to util.*stream_spec
• caps.filters() outputs FilterSummary named tuple
• caps.filter_info() outputs FilterInfo named tuple
• caps.filter_info() outputs' inputs and outputs fields returns None if dynamic and [] if none
• caps.filters() outputs' num_inputs and num_outputs fields returns None if dynamic and 0 if none
• FilterGraph class is now redirected to the new filtergraph.Graph
• improved audio.create(), image.create(), and video.create()
• configure.build_basic_vf() to use filtergraph.Graph and can be appended to user specified vf
• analyze module, updated to use filtergraph.Graph
• analyze.MetadataLogger: changed filter_spec property to filter to return a Filter object
• moved utils.error module to errors
• path.where() to raise FFmpegNotFound

• caps.filters() - fixed reporting incorrect # of pads for source and sink filters
• utils.parse_stream_spec() - fixed file_index handling
• utils.parse_stream_spec() - fixed pid stream (#) regex
• utils.filter - fixed filter and filtergraph parsing bugs

• analysis module to extract the frame metadata set by video/audio analysis filters
• video.detect() & audio.detect() to run the preset analyses
• extra_inputs argument to audio.write, image.write, video.write, and SimpleWriterBase constructor to support extra source urls
• path.check_version() to check FFmpeg version easily

• util.concat: Moved FFConcat.options to FFConcat.File. Removed FFConcat.add_option() and FFConcat.add_options()

• Support for gray10le, gray12le, and gray14le pixel formats
• ffmpegio.FFConcat.add_glob() method

• SimpleReaderBase.read() to return None if no frames remains

• BUG: SimpleWriterBase.__next__() returns empty frame at the eof
• BUG: SimpleVideoReader and SimpleAudioReader overwrites user specified options
• BUG: SimpleWriterBase fails to capture all the FFmpeg log lines

• video writers hanging when a basic filter to remove alpha channel is used

• SimpleOutputStreams output rates are now explicitly matched to input rates by default

• open() handling of write mode

• probe outputs Fraction for compatible entries

• transcode() allowed to take multiple inputs or outputs

• ffmpegio.ffmpeg and ffmpegio.ffprobe functions to call ffmpeg & ffprobe, respectively, without any input argument processing
• devices submodule to provide a framework to enumerate I/O devices
• New plugins hooks: device_source_api and device_sink_api
• plugins/devices/dshow to support Windows' DirectShow device (-f dshow)
• FFConcat class to support -f concat demuxer and ffconcat auto-scripting
• probe.frames to retrieve frame information

• Fixed failed import ffmpegio when it cannot find FFmpeg binaries
• Many minor bugs

• Uses -hwaccel auto as default for all inputs (experiment)
• Renamed ffmpeg module to path
• Refactored probe.py

• Fixed path.find(ffmpeg_dir) call

• Renamed path.get_ffmpeg to path.where; original, path.where dropped

• Fixed import failure if FFmpeg is not found in the system

• Improved FFmpeg error reporting with FFmpegError exception
• caps functions nwo throw FFmpegError exceptions
• FFmpeg is called with -nostdin flag by default

• ffmpegio.ffmpegprocess.FLAG alias for None to make flag options more readable in ffmpeg_args dict

• Fixed caps ffmpeg calls to print the banner

• finder plugin hook to allow custom automatic detection of ffmpeg executables

• Major refactoring of supporting modules, including new _utils and path modules

• -n (no overwrite) global option is now used by default. Not having this caused the subprocess to hang, waiting for user input.

• Fixed pix_fmt and pix_fmt_in check in transcode function

• Fixed option handling bugs in media readers

• All readers to support lavfi source filter inputs

• Output image dimension (v0.3.0 introduced a bug to return image as video frames)

• bytes_to_video and bytes_to_audio plugin hooks to take additional argument squeeze

• Python distribution package name from ffmpegio to ffmpegio-core
• All raw data I/O operations are performed without NumPy array (NumPy support has been spun off to python-ffmpegio-numpy repo)
• audio.create() to return sampling rate and samples
• ffmpegprocess.run() to return its stdout output as bytes
• Updated AVI processing
• Cleaned up SimpleStreams.py
• All stream classes returns rate or rate_in (instead of sample_rate or frame_rate)

• Introduced plugin system (depending on pluggy package)

• Removed acknowledging planar PCM sample_fmt's

• added 'rvv' & 'raa' open modes
• added secret bool option "_force_basic_vf"

• fixed: added media module to the root module
• fixed: exec() no longer sets stdin & stdout when unspecified
• fixed parse() handling of help options

• ffmpeg.exec() capture_log option: None = console display, False = DEVNULL

• added media.read()
• added streams.SimpleStreams.SimpleFilterBase
• added streams.SimpleStreams.SimpleAudioFilter
• added streams.SimpleStreams.SimpleVideoFilter
• added streams.AviStreams.AviMediaReader
• added 'vaf' and 'f' mode support for open()
• added ffmpegprocess.run_two_pass() for 2-pass encoding
• added 'two_pass', 'pass1_omits' and 'pass1_extras' arguments to video.write() and transcode()
• added threading.ReaderThread

• moved and renamed ffmpeg.ProgressMonitor to threading.ProgressMonitorThread
• moved and renamed utils.log.Logger to threading.LoggerThread
• changed writer argument names

• removed io.py and its dependencies

• square_pixels options to video & image
• overwrite option to video and image stream writers
• ffmpegprocess.Popen to resolve misplaced global options

• Added basic video filter options, incl. special handling of s output argument and autoconversion from transparent to opaque pix_fmt with fill_color
• Added overwrite argument to writer functions/classes
• ffmpegprocess submodule transports PIPE and DEVNULL from subprocess

• audio.create aframe argument replaced with t_in
• show_log argument displays the log at the end even if log is captured
• Renamed caps.pixfmts() and caps.samplefmts() to pix_fmts() and sample_fmts(), respectively
• Split caps.coders() to caps.decoders() and caps.encoders()
• Several output dict key names changed in caps functions

• Removed progress argument from all functions in image submodule
• Removed utils.is_forced()

• Fixed caps._getCodecInfo parsing issues

• This file CHANGELOG
• SimpleStreams.SimpleReaderBase.blocksize property to specify the number of blocks of media data to read as iterator

• Turned SimpleStreams.SimpleReaderBase.readiter() into __iter__() and __next__() to make the class Iterable
• Moved Github repo from tikuma-lsuhsc to python-ffmpegio

• Exception handling in transcode.transcode()

• First beta release.
• Main functionality of transcode, video, audio, image, SimpleStreams, probe, and caps modules.
• Preliminary implementations of FilterGraph and FFmpegError classes.