MLT Multimedia Framework

Recent News

WebVfx v1.2.0 released

Feb 16, 2020

You can download it from GitHub.

This version adds support for MLT v6.20.0’s preview scaling.

For the MLT integration, if the resource starts with “plain:” webvfx does automatic scaling by changing the zoom level. This works fine for content that is simply produced or overlayed by the filter. However, content that uses the webvfx JavaScript extension will not operate with zoom unless the service property mlt_resolution_scale is set to 1. New properties mlt_profile_scale_width and mlt_profile_scale_height are real numbers that reflect the new MLT scale properties used to do low-res preview scaling. Content script can access this via webvfx.getNumberParameter() to scale its parameters and elements.

v6.20.0 released

Feb 16, 2020

You can download it from GitHub.

This version adds support for low resolution Preview Scaling and adds a module based on the Rubber Band Library for audio pitch-shifting. An official docker image is now available on Docker Hub as mltframework/melt.

Framework

• Added consumer scaling:
  • mlt_profile_scale_width()
  • mlt_profile_scale_height()
  • Mlt::Profile::scale_width()
  • Mlt::Profile::scale_height()
  • support for a double scale property to melt and the xml producer
• Fixed mlt_properties_set() with an invalid expression.
• Added new functions that do not evaluate expressions:
  • mlt_properties_set_string()
  • Mlt::Properties::set_string()
• Improved the service-caching heuristic in mlt_multitrack.
• Fixed possible crashes in mlt_playlist get_frame() and mlt_filter_process().

Modules

• Added the rubberband module with a rbpitch filter.
• Added pitch compensation to timewarp producer.
• Added the invert_scale property to the affine filter and transition.
• Added the reverse property to the shape filter.
• Added support for text keyframes to the text and qtext filters.
• Added support for the CSRT and MOSSE algorithms in opencv.tracker filter.
• Fixed a crash on empty algo property in the opencv.tracker filter.
• Changed vorbis module to no longer be deprecated.
• Improved colorspace conversions in the avformat module.
• Fixed audio artifacts on initial seek to in point in avformat producer.
• Fixed the colorspace of the cached image in avformat producer.
• Fixed white video flashes on property changes in the qtext filter.
• Fixed a crash in the rotoscoping filter with large spline deviations.
• Fixed a crash in the sdi consumer if the driver is not loaded.
• Improved support for a video clip as luma producer to the luma transition.
• Fixed a crash in the matte transition.
• Fixed a crash when using invert property =1 in the composite transition.

Other

• Added a Dockerfile and integrated docker build into Travis CI.
• Added more avformat consumer presets:
  • intermediate/DNxHR-HQ
  • intermediate/ProRes HQ
  • ALAC
  • FLAC
• Fixed some parameters in the XDCAM and D10 avformat presets.
• Fixed link failure on some CPU architectures.

v6.18.0 released

Nov 11, 2019

You can download it from GitHub.

This version is a general maintenance release with a bunch of fixes, improvements, and additions.

Framework

• Fixed some data races in mlt_consumer, mlt_deque, and mlt_property.
• Fixed the mlt_events listener incorrect owner argument.
• Added support for the LC_ALL environmant variable on Windows.
• Fixed the argument to mlt_factory_init() not working on Windows.
• Fixed mlt_service_identify() not reliable in some use cases.
• Added some default and copy constructors and assignment operators to mlt++
  • Filter()
  • Filter( const Filter &filter )
  • Filter& operator=( const Filter &filter )
  • Producer( const Producer &producer )
  • Producer& operator=( const Producer &producer )
  • Properties( const Properties &properties )
  • Properties& operator=( const Properties &properties )
  • Service( const Service &service )
  • Service& operator=( const Service &service )
  • Transition()
  • Transition( const Transition &transition )
  • Transition& operator=( const Transition &transition )
• Added mlt_luma_map:
  • mlt_luma_map_init
  • mlt_luma_map_new
  • mlt_luma_map_render
  • mlt_luma_map_from_pgm
  • mlt_luma_map_from_yuv422
• Fixed preset overrides depend on the XML attribute order.
• Fixed serializing an animated property with a new length.

Modules

• Fixed interpolation in rotoscoping filter.
• Fixed crop filter not working with color producer.
• Fixed some data races in the sdl and sdl2 consumers.
• Fixed some data races in the avformat producer.
• Added a movit.flip filter to the opengl module.
• Fixed using filters on frei0r producers.
• Added support for in and out attributes on the consumer xml element.
• Fixed using an in point with the multi consumer.
• Fixed avfilter fails if the image size changes.
• Fixed showing superfluous decimals for seconds in the timer filter.
• Stop serializing an invalid producer as an “INVALID” text producer in xml.
• Fixed an access violation crash in wave filter.
• Added the meta.media.color_range property to the avformat producer.
• Fixed full range yuv422p not converted correctly in the avformat producer.
• Fixed the text filter not working with pango.
• Fixed a regression using dynamictext with pango.
• Added a position property to avfilter for filters that need position info.
• Fixed avfilter.subtitles not using the source position.
• Added an analyze property to vidstab filter. When set, analysis only starts and the results file written if true.
• Fixed crash combining the affine filter with the shape filter.
• Added interlace detection from AVCodecContext.field_order.
• Changed the avformat producer to not use the rescale.interp frame property. Previously, when reacale.interp == “nearest”, it would relax seeking. Now, seek accuracy is reduced during trick play (rewind or fast forward).
• Fixed swscale flags for auto-inserted scalers in avfilter.
• Fixed a double free crash in ladspa filter on channel count mismatch.
• Refactored the composite and luma transitions to use mlt_luma_map.
• Refactored the pgm producer and shape filter to use mlt_luma_map.
• Refactored the lumas module to use mlt_luma_map.
• The lumas module is now disabled by default and must be explicitly enabled.
• Added property animation to the threshold filter.
• Added a cairoblend_mode filter to the frei0r module to affect a frei0r.cairoblend transition used to composite/blend tracks.
• Added support for new vaapi options to the avformat consumer:
  • connection_type: x11 or drm
  • driver
  • kernel_driver
• Fixed the timewarp producer with a colon in the filename.
• Fixed a relative file name with a colon in it in the xml producer.
• Fixed defaulting to album or poster art if there is another video stream.
• Fixed parameter animation in frei0r plugins when using frame threads. This change also enables frame-threading for more plugins.
• Improved the qtblend filter to not process alpha if no transparency.
• Added a background_color property to the qtblend filter.
• Fixed the opencv.tracker incorrect behavior on cut clips.
• Changed opencv.tracker to store absolute frame numbers.
• Fixed incorrect frame offset on render in opencv.tracker.
• Add an alpha_over property to luma transition. This addresses a behavior regression in version 6.14.0.
• Fixed noimagecache not working in the avformat producer.

Other

• Mlt++ now requires C11 compiler support.
• Fixed closing melt SDL2 window from window manager (i.e. close button).
• Added -repository option to the melt command.
• Added unit tests for Mlt::Event.
• Fixed returning image data for Python 3.
• Switch to python3 by default.
• Updated the prores encoding presets to set vendor ID and colr atom.
• Added a CMake build system. This is not yet prefered over the existing configure script and Makefiles and has less flexibility. It is a start and has limited support.

WebVfx v1.1.0 released

Jun 15, 2019

You can download it from GitHub.

MLT continues to maintain and extend the WebVfx add-on that lets one use HTML and Qt QML technologies for a video asset or effects. Back in November, 2018 we released v1.0.0 as it has proven fairly stable and useful in Shotcut for a few years. Here are the enhancements for v1.1.0:

• Added support for alpha channels in MLT video for input and output.
• Added a transparent property to the MLT filter. When the property is set to 1, this gives a filter a transparent white canvas for drawing while the MLT video image is still available through the JavaScript extension. This facilitates using WebVfx to draw the alpha channel independently of the MLT alpha channel.
• Added support for getting a MLT rectangle property (mlt_rect) as a JavaScript object with properties: x, y, width, height, and opacity.

v6.16.0 released

May 7, 2019

You can download it from GitHub.

This version is released to facilitate packaging the latest version of Shotcut, which is using new APIs.

Framework

Added functions to get/set a creation date to a producer:

• mlt_producer_get_creation_time()
• mlt_producer_set_creation_time()
• Mlt::Producer::set_creation_time()
• Mlt::Producer::get_creation_time()

Modules

• Fixed dance filter not showing when lower track is transparent.
• Refactored dynamictext filter to use mlt_producer_get_creation_time().
• Marked frei0r rgsplit0r plugin version < 1.1 as not thread-safe.
• Fixed possible null pointer crash in mlt_properties_serialise_yaml().

Search

Subscribe

Subscribe to Newsvia RSS.