MLT Multimedia Framework

Recent News

v7.0.1 released

May 15, 2021

You can download it from GitHub.

This version is just build fixes for the most immediate problems with the somewhat new but exclusive build system in v7.

• Fixed docker image not working.
• Fixed a system-installed build cannot finds its modules and data.
• Fixed the python installation path for binaries.
• Added support for the DESTDIR environment variable when creating melt symlink.
• Increased the build constant for the maximum size of a line of a properties file.
• Fixed the vid.stab metadata install path.

v7.0.0 released

May 2, 2021

You can download it from GitHub.

This is a major new version that breaks API to add a major new feature to the framework: retiming. This is accomplished through new classes mlt_chain and mlt_link. And since we are breaking API we decided to clean house by removing deprecations and switching the build system over entirely to CMake. For more information see our migration guide. More documentation about the new chain and link classes is still pending.

Framework

• Added mlt_chain and Mlt::Chain classes.
• Added mlt_link and Mlt::Link classes.
• Added a link value to service type in the service metadata schema.
• Added a boolean animation parameter attribute to the service metadata schema.
• Added mlt_animation_shift_frame() and Mlt::Animation::shift_frames().
• Added mlt_animation_get_string().
• Fixed using a stale cached property animation string.
• Added mlt_image and Mlt::Image classes.
• Remove legacy “height + 1” workaround in image allocation.
• Fixed a crash on setting timewarp speed higher than 23x.
• Added mlt_audio_silence().
• Removed mlt_image_opengl.
• Replaced variadic arguments in mlt_events with new mlt_event_data APIs.
• Removed mlt_geometry APIs.
• Renamed mlt_image_rgb24a as mlt_image_rgba.
• Renamed mlt_image_rgb24 to mlt_image_rgb.
• Renamed mlt_image_glsl to mlt_image_movit.
• Renamed mlt_image_glsl_texture to mlt_image_opengl_texture.
• Removed virtual function mlt_frame::get_alpha_mask().
• Removed mlt_frame_get_alpha_mask().
• Removed deprecated functions:
  • mlt_sample_calculator
  • mlt_sample_calculator_to_now
  • mlt_channel_layout_name
  • mlt_channel_layout_id
  • mlt_channel_layout_channels
  • mlt_channel_layout_default
  • mlt_slices_init
  • mlt_slices_close
  • mlt_slices_run
  • mlt_playlist_move_region
  • Mlt::Playlist::move_region
• Fixed a rounding error calculating display aspect ratio in mlt_profile_from_producer().

Modules

• Added a timeremap link to the core module with animatable map property. (Speed can increase or decrease between keyframes including reverse.)
• Added chain and link XML elements to xml module.
• Added “meta.media.has_b_frames” property to avformat producer.
• Removed deprecated modules:
  • dv
  • gtk2 (not gdk)
  • kino
  • linsys
  • lumas
  • motion_est
  • swfdec
  • videostab
• Removed the following services:
  • data_feed filter
  • data_show filter
  • region filter and transition
  • sdl_image
• Converted filters to use new mlt_image class:
  • brightness
  • imageconver
  • mirror
  • spot_remover
• Deprecated the audiowave filter.
• Added the ability to build the jackrack module without JACK to get only LADSPA producers and filters.
• Deprecated start and end properties for the following filters:
  • brightness
  • panner
  • boxblur
  • wave
  • volume
• Removed deprecated font property from pango producer.
• Improved album art (attached pic) detection in avformat producer.
• Improved the resample filter to have less artifacts and use less memory.

Other

• CMake: nearly complete rewrite.
• Removed the old configure bash scripts and Makefiles.
• Added -chain and -link options to melt command line.

v6.26.1 released

Apr 10, 2021

You can download it from GitHub.

This is the last planned release of major version 6. Version 7 will be released soon and introduce some minor API breakage while removing deprecations. The main new feature in this version is hardware-accellerated decoding! However, this is a basic implementation: It always returns the uncompressed video to the CPU memory with no pipelining to filters. Even when coupled with hardware encoding in the avformat consumer it must transfer the video. Also, there is no automatic software/CPU fallback and no resource management.

Modules

• Added support for hwaccel query string parameter to the avformat producer. It accepts the following values:
  vaapi (Linux/BSD), cuda (Linux), videotoolbox (macOS), d3d11va (Windows), dxva2 (Windows)
  The query string delimiter is a backslash followed by a question mark: \?. The backslash prevents interpreting the question mark from being a part of the file name. You must preface the file name with either avformat: or file: to make it work with melt (but not XML or via API because melt interprets name=value as a MLT property if there is no colon before equals).
• Added support for hwaccel_device query string parameter to the avformat producer. This is only used with vaapi (device path) and cuda, d3d11va, or dxva (number).
• Improved the usage of image slice threading in frei0r. This only applies when threads=0 and only works with some frei0r plugins that you must decide yourself.
• Added an ellipse item to kdenlivetitle producer.
• Added support for PNG and GIF as album art in the avformat producer.
• Added BT.2020 color space metadata to the avformat producer.
• Resolved many FFmpeg deprecations in the avformat producer making it possible to support AV1 decoding.
• Added a strobe fitler that periodically makes the alpha channel transparent.
• Added a new typewriter text filter (currently only works with the kdenlivetitle producer).
• Improved sound quality for lower pitch shifts in rbpitch.
• Fixed speed of trick play in the jack, rtaudio, sdl_audio, and sdl2_audio consumers.
• Fixed matrix for independent channels in swresample filter.
• Fixed leading zeros for the timer filter.
• Fixed flickering using affine with a luma transition.
• Fixed a crash using RGBA images in the qimage producer (regression in v6.22.0).
• Fixed brightness filter misbehaves on alpha > 1.
• Fixed writing flac format file does not set its duration in the avformat consumer.
• Fixed an infinite loop in rbpitch filter.
• Fixed ttl in the qimage producer.
• Fixed building with OpenCV 4.5
• Fixed artifacts with multiple HTML qtext filters and frame threading.
• Deprecated the start and end properties on the following (use property animation instead):
  • brightness
  • panner
  • boxblur
  • wave
  • volume
• Deprecated the following services:
  • data_show
  • region
  • transition filter
  • autotrack_rectangle
  • motion_est
  • slowmotion

Other

• CMake:
  • Fixed building without SWIG.
  • Added many “MOD_…” options to explictly disable modules.
  • Added src/tests and the option BUILD_TESTING, which defaults off.
  • All dependency checks moved to top level CMakeLists.txt.
  • Install melt man page.
  • Install oldfilm SVG files.
  • Added src/examples.
  • Install framework/metaschema.yaml.
  • Fixed plusgpl datadir.
  • Added all swwig/ languages.
  • Increased C++ standard to C++14.
• Added an AV1 encoding preset.
• Improved documentation of the requirement for C11.
• The minimum version of FFmpeg is v4.0 and Libav is no longer supported.

v6.24.0 released

Dec 6, 2020

You can download it from GitHub.

This version fixes bugs associated with the preview scaling introduced in the previous version.

This version is mostly fixes plus a few new filters.

Framework

• Trigger a property-changed event on mlt_properties_pass_list.
• Fixed using a video transition with a video clip on an audio track.
• Reduce the amount of service caching to 2X #tracks to reduce memory usage.

Modules

• Added the pillar_echo filter to the plus module.
• Added a qtcrop filter to the qt module.
• Added html, resource, overflow-y, and _hide properties to the qtext filter for rich text.
• Added the filter choppy to the core module.
• Added slice threading to the brightness filter.
• Fixed compiling with OpenCV 4.
• Fixed the colors when using mlt_image_format=rgb24a with avformat consumer.
• Fixed using WebVfx in a Docker container.
• Fixed a possible crash in the timewarp producer on sources with non-integer frame rates.
• Fixed a regression in version 6.22 with multiple affine filters at the same time.
• Fixed possible abort or deadlock on recursive pthread mutexes in avformat producer.
• Fixed a crash in crop filter with large center_bias value when use_profile is 1.
• Fixed a white video frame appearing on threaded rendering in freeze filter.
• Fixed MLT XML DRD to permit empty playlists, which may occur on empty tracks in a multitrack.
• Fixed initializing QApplication in the qimage producer.
• Fixed interpolation when scaling with the affine rect and geomety properties.
• Fixed high memory usage with high factors of pitch shifting in the rbpitch filter.
• Fixed a crash on files with more than 32 streams in the avformat producer.

Other

• Fixed CMake build on MSYS2 and Windows Craft.
• Added the Python binding to the CMake build.
• Added the sdl (v1) module to the CMake build.
• Removed minrate and maxrate from the webm avformat consumer preset.

v6.22.1 released

Jul 30, 2020

You can download it from GitHub.

This version fixes bugs associated with the preview scaling introduced in the previous version.

Framework

• Added mlt_properties_exists() and Mlt::Properties::property_exists().
• Added mlt_audio C class with:
  • mlt_audio_new()
  • mlt_audio_close()
  • mlt_audio_set_values()
  • mlt_audio_get_values()
  • mlt_audio_alloc_data()
  • mlt_audio_calculate_size()
  • mlt_audio_plane_count()
  • mlt_audio_plane_size()
  • mlt_audio_get_planes()
  • mlt_audio_shrink()
  • mlt_audio_reverse()
  • mlt_audio_copy()
  • mlt_audio_calculate_frame_samples()
  • mlt_audio_calculate_samples_to_position()
  • mlt_audio_channel_layout_name()
  • mlt_audio_channel_layout_id()
  • mlt_audio_channel_layout_channels()
  • mlt_audio_channel_layout_default()
• Added Mlt::Audio C++ class with:
  • Mlt::Audio::Audio()
  • Mlt::Audio::Audio(mlt_audio_s*)
  • Mlt::Audio::~Audio()
  • Mlt::Audio::data()
  • Mlt::Audio::set_data(void*)
  • Mlt::Audio::frequency()
  • Mlt::Audio::set_frequency(int)
  • Mlt::Audio::format()
  • Mlt::Audio::set_format(mlt_audio_format)
  • Mlt::Audio::samples()
  • Mlt::Audio::set_samples(int)
  • Mlt::Audio::channels()
  • Mlt::Audio::set_channels(int)
  • Mlt::Audio::layout()
  • Mlt::Audio::set_layout(mlt_channel_layout)
• Fixed drop-frame timecode for 59.94 fps.
• Fixed crash on null pointer passed to mlt_consumer_stop().

Modules

• Fixed frei0r transitions with preview scaling.
• Fixed affine ox and oy properties incorrect with preview scaling.
• Fixed a crash and incorrect preview scaling with more than one affine filter active on the same frame.
• Fixed preview scaling for the rotoscoping filter.
• Moved some services from gtk2 module to a new gdk module:
  • gtkrescale filter
  • pango producer
  • pixbuf producer
• Deprecated the gtk2 module and no longer enabled by default.
• Deprecated the linsys (DVEO SDI) module.
• Added the sample_fmt property to the avformat consumer.
• Fixed a possible segfault in the mix transition.
• Removed support for text keyframes to the text and qtext filters to fix regression on strings containing ‘=’.
• Disable frame-threading with bigsh0t, distort0r, and medians frei0r plugins.
• Added meta.media.%d.stream.projection property the avformat producer.
• Fixed a crash with with filters not supporting preview scale in frei0r transitions.
• Fix artifacts in luma transition and affine filter with frame-threading.
• Stop including title="Anonymous Submission" in xml consumer.
• Fixed a crash in opencv.tracker filter.
• Fixed a crash in composite transition if luma file fails to load.
• Added validations in opengl module to prevent asserts in Movit.
• Fixed building with OpenCV 4.
• Changed avformat producer to accept a ‘?’ in argument/resource property by escaping it as ‘\?’.
• Changed the background property of the affine filter to be mutable.
• Fixed changing the audio_index property in the avformat producer.
• Changed resample filter to more resiliant to frequency changes.
• Added a video_delay property to the sdl2_audio and rtaudio consumers.
• Add millisecond options to the timer filter.
• Fixed the in point handling for the timewarp producer.
• Fixed some audio gaps and sync issues with the rbpitch filter and timewarp pitch compensation.
• Fixed a possible crash caused by producer consumer.
• Changed avformat consumer to set AVOption color_primaries based on the MLT colorspace if not already set as property.
• Fixed crop right on image with odd width skews image in crop filter.
• Fixed incorrect silence value for unsigned 8-bit audio in avformat producer.
• Changed qimage to use Qt’s internal orientation detection instead of libexif.
• Reduced clicks in mix transition by silencing buffers on discontinuity.
• Improved A/V synchronization in swresample and resample filters - also reduces audio clicks.
• Improved speed of the qimage producer.
• Fixed incorrect color using libx264rgb in avformat consumer.
• Fixed relative paths for avfilters that have the filename option.
• Fixed some avfilters dropping the alpha channel: smartblur, vaguedenoiser.
• Improved performance of the resize filter.
• Fixed an affine filter inside a transition was always nearest neighbor interpolation.
• Changed the lift_gamma_gain filter to use round values up.

Other

• Fixed melt option -group applies to an implicit consumer.
• Added -quiet option to melt (implies -silent but more so).
• CMake build improvments adding modules:
  • gdk
  • jackrack
  • lumas
  • resample
  • sox
  • vorbis
• Added avformat consumer presets:
  • Slide-Deck-H264
  • Slide-Deck-HEVC
• Removed “intra=1” from some avformat presets (use g=1 for intra only):
  • intermediate/MPEG-2
  • intermediate/MPEG-4
  • lossless/H.264
• Fixed using Qt, Movit, and WebVfx in the official docker image:
  https://hub.docker.com/repository/docker/mltframework/melt
  IMPORTANT: it now requires docker run with the --init option.

Search

Subscribe

Subscribe to Newsvia RSS.