All of this is speculative and only ideas for consideration at this point.
7.x (Current)
See also Version 7 Migration Guide
- convert most services and properties to
mlt_animation
- add more tests (difficult to define coverage required at this point)
- document expectations in each type of service (e.g. meta properties, frame-threading)
- improve
mlt_consumer:real_time > 1
- letter spacing in text services
- fix the cache issue that necessitates
mlt_multitrack.c:resize_service_caches()
8.x (Next)
- add const and cleanup constructors in C++ API details TBD
- add option to build without locale
mlt_profile -> mlt_context derived from mlt_properties, namely for reference-counting but also to replace mlt_environment()- add color_trc, color_range, color_primaries, interpolation, and audio params (channels, frequency, channel_layout) to
mlt_context
- tighten the property namespace (inconsistent service name prefixing, hiding, and putting things on
mlt_frame)
- add
mlt_rational type and property functions, and convert things like aspect ratios and frame rates to use it
- add 16-byte alignment for FFmpeg including aligned image plane and audio channel buffers
- change
mlt_frame_get_image() to take mlt_image instead of multiple parameters (as well as get_image in all services)
- change
mlt_frame_get_audio() to take mlt_audio instead of multiple parameters (as well as get_audio in all services)
- collapse
mlt_tractor, mlt_multitrack, and mlt_field APIs
- speed < 1.0 support (
mlt_position as double?)
- variable frame-rate support
- add a way for services to mark frames as dirty to faciliate a disk cache or smart render (pass-through coded packets)
- builds with MSVC and vcpkg (see its repo)
- port many effects to Movit or a new image processing framework that:
- supports 10-bit input and output (future support for 12-bit desired)
- supports HDR (HDR10/PQ, HLG)
- maximizes usage of multi-core CPU and GPU
- ensures scene-referred linear color image processing
- compatible with avfilter, OpenCV, and Qt high bit depth
- determine new
mlt_image_formats needed for new processing framework
- integrate OpenColorIO
- add OpenTimelineIO producer and consumer
- automate and test arm64 builds
- add an abstract playback consumer that encompasses the logic of
sdl2 but only fires events and does not integrate
any technologies outside of MLT
- add support for timed text
- add support for timed data
- find a framework to integrate AI/ML-based effects
- background removal
- upscaling
- denoising
- sharpening
- optical flow
