After nearly a year of development, GStreamer 1.28 is now available. It's a cross-platform set of components for creating a wide range of multimedia applications, from media players and audio/video file converters to VoIP applications and streaming systems. GStreamer's code is licensed under the LGPLv2.1 license. Separate updates are being developed for the gst-plugins-base, gst-plugins-good, gst-plugins-bad, and gst-plugins-ugly plugins, as well as the gst-libav framework and the gst-rtsp-server streaming server. The new release is backwards compatible with the 1.0 branch at the API and ABI level. Binary builds will soon be prepared for AndroidIOS, macOS и Windows (In Linux (It is recommended to use packages from the distribution).
Key improvements in GStreamer 1.28:
- A new HIP (Heterogeneous-computing Interface for Portability) plugin has been added to support AMD's heterogeneous computing API, which uses GPU parallel computing capabilities in addition to CPUs. HIP simplifies the implementation of cross-GPU support within a single codebase (using the ROCm stack for execution on AMD GPUs, and a CUDA translation layer for NVIDIA GPUs). The plugin provides components implemented using HIP for operations such as video mixing, color space conversion, video resizing, and data exchange between system and video memory. A library has also been developed for integrating HIP support into applications.
- Support has been added for the LCEVC (Low Complexity Enhancement Video Coding) video coding standard, which adds an additional metadata layer on top of the standard H.265 and H.266 codecs to improve video quality. The LCEVC encoder and decoder are based on libraries from the V-Nova SDK.
- Added support for Vulkan graphics API capabilities to accelerate AV1 and VP9 video decoding, as well as H.264 video encoding. The Vulkan-based H.265 video decoder now supports 10-bit per channel color depth.
- The gtkwaylandsink element, designed for rendering using GTK4 and Wayland, has expanded its color capabilities, such as adding support for parsing and setting HDR10 metadata.
- A new element for separating audio sources (for example, separating voices from background music) has been added, based on the charon-audio library, which implements the Demucs method in Rust. Additionally, a Deepgram speech recognition plugin written in Rust is included.
- Analytics and object recognition capabilities have been expanded. Elements for combining and separating analytics have been added. The tensor-decoder classifier, the facedetector face detector, the YOLOv8 object detector, and an execution element and decoder for the YOLOX object recognition model have been implemented. A tensordecodebin element for automatically connecting decoders and an element for executing AI models using LiteRT (a replacement for TensorFlow Lite) have been added.
- Added a new element for decoding GIF images, written in Rust and supporting looping animation.
- The input-selector element implements a two-phase mechanism for switching between input streams (sinkpad), eliminating the occurrence of race conditions when switching.
- The webrtcsink element has been updated to support hardware-accelerated encoding using the VA-API and the ability to renegotiate connections.
- Added elements for combining and extracting streams in the ST-2038 broadcast standard format.
- Added support for encoded streams to the fallbacksrc element.
- The FLV media container used for RTMP broadcasting has improved support for H.265 video and added support for multi-channel audio.
- The glupload element has been updated to support a udmabuf-based uploader, allowing buffer sharing between software decoders/sources and GPUs, display engines (wayland), and other devices.
- Added qml6 element for rendering using Qt6 QML.
- The video library now includes operations for cropping, scaling, rotating, and shifting.
- Added a GstContext object to allow the use of a shared thread pool when executing elements to improve the performance of operations such as compositing and video conversion.
- Added icecastsink element for receiving streams from IceCast, supporting AAC format and written in Rust.
- Added new ElevenLabs voice cloning element.
- A mode for accelerating playback while preserving pitch has been added to the speech synthesizer.
- Added a vmaf element to assess the perceived quality of video using Netflix's VMAF framework.
- Added a new MPEG-H audio decoding plugin based on the Fraunhofer MPEG-H decoder and supporting unpacking of MP4 media containers.
- Expanded support for closed captions (CC) and text processing.
Source: opennet.ru
