gstreamer/subprojects/gst-plugins-bad/NEWS

GStreamer 1.26 Release Notes

GStreamer 1.26.0 was originally released on 11 March 2025.

See https://gstreamer.freedesktop.org/releases/1.26/ for the latest version of this document.

Last updated: Tuesday 11 March 2025, 20:00 UTC (log)

## Introduction

The GStreamer team is proud to announce a new major feature release in the stable 1.x API series of your favourite
cross-platform multimedia framework!

As always, this release is again packed with many new features, bug fixes, and other improvements.

## Highlights

-   H.266 Versatile Video Coding (VVC) codec support
-   Low Complexity Enhancement Video Coding (LCEVC) support
-   Closed captions: H.264/H.265 extractor/inserter, cea708overlay, cea708mux, tttocea708 and more
-   New hlscmafsink, hlssink3, and hlsmultivariantsink; HLS/DASH client and dashsink improvements
-   New AWS and Speechmatics transcription, translation and TTS services elements, plus translationbin
-   Splitmux lazy loading and dynamic fragment addition support
-   Matroska: H.266 video and rotation tag support, defined latency muxing
-   MPEG-TS: support for H.266, JPEG XS, AV1, VP9 codecs and SMPTE ST-2038 and ID3 meta; mpegtslivesrc
-   ISO MP4: support for H.266, Hap, Lagarith lossless codecs; raw video support; rotation tags
-   SMPTE 2038 ancillary data streams support
-   JPEG XS image codec support
-   Analytics: New TensorMeta; N-to-N relationships; Mtd to carry segmentation masks
-   ONVIF metadata extractor and conversion to/from relation metas
-   New originalbuffer element that can restore buffers again after transformation steps for analytics
-   Improved Python bindings for analytics API
-   Lots of Vulkan integration and Vulkan Video decoder/encoder improvements
-   OpenGL integration improvements, esp. in glcolorconvert, gldownload, glupload
-   Qt5/Qt6 QML GL sinks now support direct DMABuf import from hardware decoders
-   CUDA: New compositor, Jetson NVMM memory support, stream-ordered allocator
-   NVCODEC AV1 video encoder element, and ndsdewarp
-   New Direct3D12 integration support library
-   New d3d12swapchainsink and d3d12deinterlace elements and D3D12 sink/source for zero-copy IPC
-   Decklink HDR support (PQ + HLG) and frame scheduling enhancements
-   AJA capture source clock handling and signal loss recovery improvements
-   RTP and RTSP: New rtpbin sync modes, client-side MIKEY support in rtspsrc
-   New Rust rtpbin2, rtprecv, rtpsend, and many new Rust RTP payloaders and depayloaders
-   webrtcbin support for basic rollbacks and other improvements
-   webrtcsink: support for more encoders, SDP munging, and a built-in web/signalling server
-   webrtcsrc/sink: support for uncompressed audio/video and NTP & PTP clock signalling and synchronization
-   rtmp2: server authentication improvements incl. Limelight CDN (llnw) authentication
-   New Microsoft WebView2 based web browser source element
-   The GTK3 plugin has gained support for OpenGL/WGL on Windows
-   Many GTK4 paintable sink improvements
-   GstPlay: id-based stream selection and message API improvements
-   Real-time pipeline visualization in a browser using a new dots tracer and viewer
-   New tracers for tracking memory usage, pad push timings, and buffer flow as pcap files
-   VA hardware-acclerated H.266/VVC decoder, VP8 and JPEG encoders, VP9/VP8 alpha decodebins
-   Video4Linux2 elements support DMA_DRM caps negotiation now
-   V4L2 stateless decoders implement inter-frame resolution changes for AV1 and VP9
-   Editing services: support for reverse playback and audio channel reordering
-   New QUIC-based elements for working with raw QUIC streams, RTP-over-QUIC (RoQ) and WebTransport
-   Apple AAC audio encoder and multi-channel support for the Apple audio decoders
-   cerbero: Python bindings and introspection support; improved Windows installer based on WiX5
-   Lots of new plugins, features, performance improvements and bug fixes

## Major new features and changes

### H.266 Versatile Video Coding (VVC) codec support

-   The H.266 / VVC video codec is a successor to H.265 / HEVC and is standardised in ISO/IEC 23090-3.

-   A new h266parse element was added, along with parsing API, typefinding support and some codec utility functions in the
    gst-plugins-base utility library.

-   A H.266 decoder base class for hardware-accelerated decoders was added and used to implement a VA-API-based
    hardware-accelerated H.266 decoder.

-   The FFmpeg H.266 decoder is exposed now (from FFmpeg 7.0 onwards).

-   H.266 / VVC muxing and demuxing support was implemented for MP4, Matroska and MPEG-TS containers.

-   A VVdeC-based H.266 decoder element was added to the Rust plugins, based on the Fraunhofer Versatile Video Decoder library.

### JPEG XS image codec support

-   JPEG XS is a visually lossless, low-latency, intra-only video codec for video production workflows, standardised in ISO/IEC
    21122.

-   JPEG XS encoder and decoder elements based on the SVT JPEG XS library were added, including support for muxing JPEG XS into
    MPEG-TS and demuxing it. Both interlaced and progressive modes are supported.

### Low Complexity Enhancement Video Coding (LCEVC) support

-   LCEVC is a codec that provides an enhancement layer on top of another codec such as H.264 for example. It is standardised as
    MPEG-5 Part 2.

-   LCEVC encoder and decoder elements based on V-Nova’s SDK libraries were added, including support in h264parse for extracting
    the enhancement layer from H.264 and decoding it via a lcevch264decodebin element.

### Closed captions improvements

-   New H.264 and H.265 closed captions extractor and inserter elements.

    -   These extractor elements don’t actually extract captions from the bitstream, but rely on parser elements to do that and
        add them to buffers in form of caption metas. The problem is that streams might contain B-frames, in which case the
        captions in the bitstream will not be in presentation order and extracting them requires frame-reordering in the same
        way that a decoder would do. These new elements will do exactly that and allow you to extract captions in presentation
        order without having to decode the stream.

    -   The inserter elements do something similar and insert caption SEIs into the H.264 or H.265 bitstream, taking into
        account frame ordering.

    -   This is useful if one wants to extract, process and re-insert captions into an existing video bitstream without decoding
        and re-encoding it (in which case the decoder and encoder would handle the caption reordering).

-   cdpserviceinject: New element for injecting a CDP service descriptor into closed caption CDP packets

-   cea708overlay: New element for overlaying CEA608 / CEA708 closed captions over video streams.

-   The cc708overlay element has been deprecated in favour of the cea708overlay element from the Rust plugins set.

-   cea608mux gained a "force-live" property to make it always in live mode and aggregate on timeout regardless of whether any
    live sources are linked upstream.

-   cea708mux: New element that allows to mux multiple CEA708 services into a single stream.

-   cccombiner has two new properties:

    -   "input-meta-processing" controls how input closed caption metas are processed and can be used to e.g. discard closed
        captions from the input pad if the matching video buffer already has closed caption on it.

    -   "schedule-timeout" to support timing out captions without EOS

-   tttocea708: New element for converting timed-text to CEA708 closed captions

-   Miscellaneous improvements and spec compliance fixes

### Speech to Text, Translation and Speech Synthesis

-   awstranscriber2, awstranslate: New elements around the AWS transcription and translation services.

-   polly: New element around the AWS text-to-speech polly services

-   speechmatics: New transcriber / speech-to-text and translation element

-   translationbin: Helper bin around translation elements, similar to the already existing transcriberbin for transcriptions.

### HLS DASH adaptive streaming improvements

-   The adaptivedemux2 client implementation gained support for file:// URIs and as such the ability to play HLS and DASH from
    local files. It also no longer sends spurious flush events when it loses sync in live streams, as that is unexpected and
    will be handled poorly in non-playback scenarios. Lastly, support for HTTP request retries was added via the "max-retries"
    property, along with some exponential backoff logic which can be fine-tuned via properties.

-   dashsink has received period duration fixes for dynamic MPDs and some memory leak fixes.

-   hlscmafsink, hlssink3: New single-variant HLS sink elements that can output CMAF (fMP4) or MPEG-TS fragments.

-   hlsmultivariantsink: New sink element that can output an HLS stream with multiple variants

### splitmuxsrc, splitmuxsink: lazy loading and dynamic fragment addition

-   splitmuxsrc and splitmuxsink were originally designed to handle a small number of large file fragments, e.g. for situations
    where one doesn’t want to exceed a certain file size when recording to legacy file systems. It was also designed for playing
    back a static set of file fragments that have been created by an earlier recording session and no longer changes. Over time
    people have found more applications and use cases for the splitmux elements and have been deploying them in different
    scenarios, exposing the limits of the current implementation.

-   In this release, splitmuxsink and splitmuxsrc gained new abilities aimed at improving support for recordings with a large
    number of files, and for adding fragments on the fly to allow playback of ongoing recordings:

    -   You can now add fragments directly to splitmuxsrc and provide the offset and duration in the stream:

        -   Providing offset and duration means splitmuxsrc doesn’t need to scan the file to measure it and calculate it. That
            makes for much faster startup.

        -   The new "add-fragment" signal can be used to add files to the set dynamically - allowing to be playing an ongoing
            recording and adding files to the playback set as they are finished.

    -   splitmuxsrc no longer keeps all files open, but instead only keeps 100 files open by default, configurable with the
        "num-open-fragments" property.

    -   There is a new "num-lookahead" property on splitmuxsrc to trigger (re)opening files a certain distance ahead of the play
        position.

    -   splitmuxsink will report fragment offset and fragment duration via a message on the bus when closing a file. This
        information can then be used to add the new fragment to a splitmuxsrc.

### MPEG-TS container format improvements

-   The MPEG-TS muxer and demuxer gained support for

    -   H.266 / VVC video muxing and demuxing
    -   JPEG-XS video muxing and demuxing
    -   VP9 video muxing and demuxing (using a custom mapping)
    -   AV1 video muxing and demuxing (using a custom mapping, since the work-in-progress specification effort doesn’t seem to
        be going anywhere anytime soon)
    -   SMPTE ST-2038 ancillary metadata streams (see section above)

-   mpegtsmux gained support for muxing ID3 metadata into the TS stream, as well as SMPTE 302M audio.

-   It’s also possible to disable sending SCTE-35 null (heartbeat) packets now in mpegtsmux by setting the
    "scte-35-null-interval" to 0.

-   tsparse now handles 192-byte M2TS packets

-   mpegtslivesrc: New source element that can wrap a live MPEG-TS source (e.g. SRT or UDP source) and provides a clock based on
    the PCR.

### Matroska container format improvements

-   H.266 / VVC video muxing and demuxing support

-   matroskamux

    -   was ported to the GstAggregator base class, ensuring defined-latency muxing in live streaming pipelines.
    -   gained rotation tag support

-   matroskademux now also supports seeks with a stop position in push mode.

### ISO MP4 container format improvements

-   can mux and demux H.266 / VVC in MP4 now

-   can demux Hap video now, as well as Lagarith lossless video and ISO/IEC 23003-5 raw PCM audio.

-   qtdemux handles keyunit-only trick mode also in push mode now

-   support for ISO/IEC 23001-17 raw video in MP4 in qtdemux and isomp4mux.

-   support for rotation tags in the muxers and demuxers was improved to correctly handle per-media and per-track rotations, and
    support for flips was added as well.

SMPTE 2038 ancillary data streams

-   SMPTE 2038 (pdf) is a generic system to put VBI-style ancillary data into an MPEG-TS container. This could include all kinds
    of metadata such as scoreboard data or game clocks, and of course also closed captions, in this case in form of a distinct
    stream completely separate from any video bitstream.

-   A number of new elements in the GStreamer Rust closedcaption plugin add support for this, along with mappings for it in the
    MPEG-TS muxer and demuxer. The new elements are:

    -   st2038ancdemux: splits SMPTE ST-2038 ancillary metadata (as received from tsdemux) into separate streams per DID/SDID
        and line/horizontal_offset. Will add a sometimes pad with details for each ancillary stream. Also has an always source
        pad that just outputs all ancillary streams for easy forwarding or remuxing, in case none of the ancillary streams need
        to be modified or dropped.

    -   st2038ancmux: muxes SMPTE ST-2038 ancillary metadata streams into a single stream for muxing into MPEG-TS with
        mpegtsmux. Combines ancillary data on the same line if needed, as is required for MPEG-TS muxing. Can accept individual
        ancillary metadata streams as inputs and/or the combined stream from st2038ancdemux.

        If the video framerate is known, it can be signalled to the ancillary data muxer via the output caps by adding a
        capsfilter behind it, with e.g. meta/x-st-2038,framerate=30/1.

        This allows the muxer to bundle all packets belonging to the same frame (with the same timestamp), but that is not
        required. In case there are multiple streams with the same DID/SDID that have an ST-2038 packet for the same frame, it
        will prioritise the one from more recently created request pads over those from earlier created request pads (which
        might contain a combined stream for example if that’s fed first).

    -   st2038anctocc: extracts closed captions (CEA-608 and/or CEA-708) from SMPTE ST-2038 ancillary metadata streams and
        outputs them on the respective sometimes source pad (src_cea608 or src_cea708). The data is output as a closed caption
        stream with caps closedcaption/x-cea-608,format=s334-1a or closedcaption/x-cea-708,format=cdp for further processing by
        other GStreamer closed caption processing elements.

    -   cctost2038anc: takes closed captions (CEA-608 and/or CEA-708) as produced by other GStreamer closed caption processing
        elements and converts them into SMPTE ST-2038 ancillary data that can be fed to st2038ancmux and then to mpegtsmux for
        splicing/muxing into an MPEG-TS container. The line-number and horizontal-offset properties should be set to the desired
        line number and horizontal offset.

### Analytics

-   Added a GstTensorMeta: This meta is designed to carry tensors from the inference element to the model-specific tensor
    decoder. This also includes a basic GstTensor class containing a single tensor. The actual tensor data is a GstBuffer.

-   Add N_TO_N relationship to GstAnalyticsRelationMeta: This makes it possible to describe N-to-N relationships. For example,
    between classes and regions in an instance segmentation.

-   Add a new analytics Mtd to carry segmentation masks: Being part of the GstAnalyticsMeta, it can be in relationship with the
    other Mtd, such as the classification and object detection bounding boxes.

-   onvifmetadataextractor: New element that can extract ONVIF metadata from GstMetas into a separate stream

-   originalbuffer: New plugin with originalbuffersave / originalbufferrestore elements that allow saving an original buffer,
    modifying it for analytics, and then restoring the original buffer content while keeping any additional metadata that was
    added.

-   relationmeta: New plugin with elements converting between GstRelationMeta and ONVIF XML metadata.

-   Improved Python bindings for a more Pythonic interface when iterating over GstRelationMeta’s mtd

### Vulkan integration enhancements

-   Vulkan Integration Improvements:

    -   Memory Management: Non-coherent memory is now invalidated when mapping for read in vkmemory.

    -   Color Space Selection: The vkswapper component now chooses the correct color space based on the format.

    -   Vulkan Plugin Compatibility: Support added for cases where glslc is not available for building Vulkan plugins, along
        with GLSL compiler support for glslang.

    -   Fullscreen Quad Updates: Improved support for setting NULL input/output buffers and added checks for unset video info.

    -   Vulkan Buffer Pool Enhancements: Buffer pool access flags and usage configurations have been refined, offering better
        performance for video decoding and encoding.

-   Decoder/Encoder Improvements:

    -   H264 Decoder: Enhancements to the vkh264dec component for better support of extended profiles and interlaced content
        decoding.

    -   H265 Decoder fixes: vkh265dec updated for proper handling of VPS/SPS on demand, along with fixes to PicOrderCntVal.

    -   Encoder Performance: Various internal optimizations to the Vulkan encoder, including removal of redundant references and
        better management of the DPB view.

-   Vulkan Instance and Device Management:

    -   Device Handling: Added new utility functions for managing Vulkan device instances, including
        gst_vulkan_instance_create_device_with_index and gst_vulkan_ensure_element_device.

    -   Device Context Management: Updates to manage Vulkan context handling more effectively within the application.

### OpenGL integration enhancements

-   glcolorconvert gained support for more formats and conversions:

    -   Planar YUV <-> planar YUV conversions
    -   Converting to and from v210 in general
    -   v210 <-> planar YUV
    -   UYVY and YUY2 <-> planar YUV
    -   v210 <-> UYVY and YUY2
    -   Support for Y444_10, Y444_16, I422_10, I422_12 pixel formats (both little endian and big endian variants)

-   gldownload can import DMABufs from a downstream pool

-   glupload gained a DRM raw uploader

### Qt5 + Qt6 QML integration improvements

-   qmlglsink, qml6glsink now support external-oes textures, which allows direct DMABuf import from hardware decoders. Both also
    support NV12 as an input format now.

-   qmlglsink gained support for RGB16/BGR16 as input format

-   qmlgl6src can now use a downstream buffer pool when available

-   qmlgloverlay make the depth/stencil buffer optional, which reduces memory bandwidth on Windows.

### CUDA / NVCODEC integration and feature additions

-   Added AV1 video encoder nvav1enc

-   CUDA mode nvcuda{CODEC}enc encode elements are renamed to nv{CODEC}enc and old nv{CODEC}enc implementations are removed

-   Added support for CUDA Stream-Ordered allocator

-   Added cudacompositor element which is equivalent to the software compositor element but uses CUDA

-   Added support for CUDA kernel precompile at plugin build time using nvcc and NVCODEC plugin can cache/reuse compiled CUDA
    CUBIN/PTX

-   cudaupload and cudadownload elements can support Jetson platform’s NVMM memory in addition to already supported desktop NVMM
    memory

-   Introduced nvdswrapper plugin which uses NVIDIA DeepStream SDK APIs with gst-cuda in an integrated way:

    -   nvdsdewarp: NVIDIA NVWarp360 API based dewarping element

### GStreamer Direct3D12 integration

-   New gst-d3d12 public library. The following elements are integrated with the gst-d3d12 library:

    -   NVIDIA NVCODEC decoders and encoders can support D3D12 memory
    -   Intel QSV encoders can accept D3D12 memory
    -   All elements in dwrite plugin can support D3D12 memory

-   The d3d12 library and plugin can be built with MinGW toolchain now (in addition to MSVC)

-   D3D12 video decoders and d3d12videosink are promoted to have higher rank than D3D11 elements

-   Added support for multiple mip-levels D3D12 textures:

    -   Newly added d3d12mipmapping element can generate D3D12 textures with multiple mip-levels

    -   max-mip-levels property is added to d3d12convert, d3d12videosink, and d3d12swapchainsink element, so that the elements
        can generate an intermediate texture with multiple mip-levels in order to reduce downscale aliasing artifacts

-   d3d12convert, d3d12videosink, and d3d12swapchainsink support the GstColorBalanceInterface to offer color balancing functions
    such as hue, saturation, and brightness adjustment

-   Added d3d12ipcsink and d3d12ipcsrc elements for zero-copy GPU memory sharing between processes

-   d3d12upload and d3d12download support direct GPU memory copy between D3D12 and D3D12 resources

-   Added d3d12swapchainsink element to support DirectComposition or UWP/WinUI3 SwapChainPanel based applications

-   Added d3d12deinterlace element which performs deinterlacing using a GPU vendor agnostic compute shader.

-   d3d12screencapturesrc element can capture HDR enabled desktop correctly in DDA mode (DXGI Desktop Duplication API)

### Capture and playout cards support

-   ajasrc: Improve clock handling, frame counting, capture timestamping, and signal loss recovery

-   The Blackmagic Decklink plugin gained support

    -   for HDR output and input (PQ + HLG static metadata)

    -   all modes of Quad HDMI recorder

    -   scheduling frames before they need to be displayed in decklinkvideosink

### RTP and RTSP stack improvements

-   rtspsrc now supports client-managed MIKEY key management. Some RTSP servers (e.g. Axis cameras) expect the client to propose
    the encryption key(s) to be used for SRTP / SRTCP. This is required to allow re-keying. This mode can be enabled by enabling
    the "client-managed-mikey-mode" property and comes with a number of new signals ("request-rtp-key" and "request-rtcp-key"),
    action signals ("set-mikey-parameter" and "remove-key") and properties ("hard-limit" and "soft-limit").

-   rtpbin: Add new “never” and “ntp” RTCP sync modes

    -   Never is useful for some RTSP servers that report plain garbage both via RTCP SR and RTP-Info, for example.
    -   NTP is useful if synchronization should only ever happen based on RTCP SR or NTP-64 RTP header extensions.

    This is part of a bigger refactoring of the synchronization / offsetting code in rtpbin, which also makes it regularly emit
    the sync signal even if no new synchronisation information is available, controlled by the new "min-sync-interval" property.

-   rtpjitterbuffer: add RFC7273 active status to jitterbuffer stats so applications can easily check whether RFC7273 sync is
    active.

-   rtph265depay: Add "wait-for-keyframe" "request-keyframe" properties and improve request keyframe logic

-   rtppassthroughpay gained the ability to regenerate RTP timestamps from buffer timestamps via the new "retimestamp-mode"
    property. This is useful in a relay RTSP server if one wants to do full drift compensation and ensure that the stream coming
    out of gst-rtsp-server is not drifting compared to the pipeline clock and also not compared to the RTCP NTP times.

-   New Rust RTP payloaders and depayloaders for AC3, AMR, JPEG, KLV, MPEG-TS (MP2T), MPEG-4 (MP4A, MP4G), Opus, PCMU (uLaw),
    PCMA (aLaw), VP8, VP9.

-   New rtpbin2 based on separate rtprecv and rtpsend elements

### WebRTC improvements

-   webrtcbin improvements

    -   Make basic rollbacks work

    -   Add "reuse-source-pads" property: When set to FALSE, if a transceiver becomes send-only or inactive then pre-existing
        source pads will receive an EOS event and no further traffic even after further renegotiation. When TRUE, pads will
        simply not receive any output when the negotiated transceiver state doesn’t have incoming traffic. If renegotiated
        later, the pad will receive data again.

    -   Early CNAME support (RFC5576): Have CNAME available to the jitterbuffer before the the first RTCP SR is received, for
        rapid synchronization.

    -   New "post-rtp-aux-sender" signal to allow for placement of an object after rtpbin, before sendbin. This is useful for
        objects such as congestion control elements, that don’t want to be burdened by the synchronization requirements of
        rtpsession.

    -   Create and associate transceivers earlier in negotiation, and other spec compliance improvements

    -   Statistics generation improvements for bundled streams

-   webrtcsink improvements:

    -   Support for more encoders: nvv4l2av1enc, vpuenc_h264 (for imx8mp), nvav1enc, av1enc, rav1enc and nvh265enc.

    -   The new "define-encoder-bitrates" signal allows applications to fine-tune the bitrate allocation for individual streams
        in cases where there are multiple encoders. By default the bitrate is split equally between encoders.

    -   A generic mechanism was implemented to forward metas over the control channel.

    -   Added a mechanism for SDP munging to handle server-specific quirks.

    -   Can expose a built-in web server and signalling server for prototyping and debugging purposes.

-   webrtcsink and webrtcsrc enhancements:

    -   Support for raw payloads, i.e. uncompressed audio and video

    -   NTP & PTP clock signalling and synchronization support (RFC 7273)

    -   Generic data channel control mechanism for sending upstream events back to the sender (webrtcsink)

-   webrtcsrc now has support for multiple producers

## New elements and plugins

-   Many exciting new Rust elements, see Rust section below.

-   webview2src: new Microsoft WebView2 based web browser source element

-   h264ccextractor, h264ccinserter: H.264 closed caption extractor / inserter

-   h265ccextractor, h265ccinserter: H.265 closed caption extractor / inserter

-   h266parse

-   lcevch264decodebin

-   New VA elements (see below): vah266dec, vavp8enc, vajpegenc, vavp8alphadecodebin, vavp9alphadecodebin

-   svtjpegxsdec, svtjpegxsenc: SVT JPEG XS decoder/encoder

-   Many other new elements mentioned in other sections (e.g. CUDA, NVCODEC, etc.)

## New element features and additions

-   audioconvert enhancements:

    -   Add possibility to reorder input channels when audioconvert has unpositionned audio channels as input. It can now use
        reordering configurations to automatically position those channels via the new "input-channels-reorder" and
        "input-channels-reorder-mode" properties.

    -   Better handling of setting of the mix matrix at run time

    -   handles new GstRequestMixMatrix custom upstream event

-   audiorate: Take the tolerance into account when filling gaps; better accounting of the number of samples added or dropped.

-   av1enc: Add "timebase" property to allow configuring a specific time base, in line with what exists for vp8enc and vp9enc
    already.

-   av1parse can parse annexb streams now, and typefinding support has been added for annexb streams as well.

-   The GTK3 plugin has gained support for OpenGL/WGL on Windows

-   fdsrc has a new "is-live" property to make it act like a live source and timestamp the received data with the clock running
    time.

-   imagefreeze: Add support for JPEG and PNG

-   kmssink: Extended the functionality to support buffers with DRM formats along with non-linear buffers

-   pitch now supports reverse playback

-   queue can emit the notify signal on queue level changes if the "notify-levels" property has been set.

-   qroverlay: the "pixel-size" property has been removed in favour of a new "size" property with slightly different semantics,
    where the size of the square is expressed in percent of the smallest of width and height.

-   rsvgdec: Negotiate resolution with downstream and scale accordingly

-   rtmp2: server authentication improvements

    -   Mimic librtmp’s behaviour and support additional connection parameters for the connect packet, which are commonly used
        for authentication, via the new "extra-connect-args" property.

    -   Add support for Limelight CDN (llnw) authentication

-   scaletempo has gained support for a “fit-down” mode: In fit-down mode only 1.0 rates are supported, and the element will fit
    audio data in buffers to their advertised duration. This is useful in speech synthesis cases, where elements such as
    awspolly will generate audio data from text, and assign the duration of the input text buffers to their output buffers.
    Translated or synthesised audio might be longer than the original inputs, so this makes sure the audio will be sped up to
    fit the original duration, so it doesn’t go out of sync.

-   souphttpsrc: Add the notion of "retry-backoff" and retry on 503 (service unavailable) and 500 (internal server error) http
    errors.

-   taginject now modifies existing tag events of the selected scope, with the new "merge-mode" property allowing finer control.

-   timecodestamper gained a new running-time source mode that converts the buffer running time into timecodes.

-   playbin3, uridecodebin3, parsebin

    -   lots of stream-collection handling and stability/reliability fixes
    -   error/warning/info message now include the URI (if known) and stream-id
    -   missing plugin messages include the stream-id

-   videocodectestsink gained support for GBR_10LE, GRAY8 and GRAY10_LE{16,32} pixel formats

-   videoflip gained support for the Y444_16LE and Y444_16BE pixel formats

-   videoconvertscale:

    -   Handle pixel aspect ratios with large numerators or denominators
    -   Explicitly handle the overlaycomposition meta caps feature, so it doesn’t get dropped unnecessarily

-   waylandsink prefers DMABuf over system memory now

-   x264enc has a new "nal-hrd" property to make the encoder signal HRD information, which is required for Blu-ray streams,
    television broadcast and a few other specialist areas. It can also be used to force true CBR, and will cause the encoder to
    output null padding packets.

-   zbar: add support for binary mode and getting symbols as raw bytes instead of a text string.

## Plugin and library moves

-   macOS: atdec was moved from the applemedia plugin in -bad to the osxaudio plugin in -good, in order to be able to share
    audio-related helper methods.

## Plugin and element removals

-   None in this cycle

## Miscellaneous API additions

### GStreamer Core

-   gst_meta_api_type_set_params_aggregator() allows setting an GstAllocationMetaParamsAggregator function for metas, which has
    been implemented for GstVideoMeta and is used to aggregate alignment requirements of multiple tee branches.

-   gst_debug_print_object() and gst_debug_print_segment() have been made public API. The can be used to easily get string
    representations of various types of (mini)objects in custom log handlers.

-   Added gst_aggregator_push_src_event(), so subclasses don’t just push events directly onto the source pad bypassing the base
    class without giving it the chance to send out any pending serialised events that should be sent out before.

-   GstMessage has gained APIs to generically add “details” to messages:

    -   gst_message_set_details()
    -   gst_message_get_details()
    -   gst_message_writable_details()
    -   gst_message_parse_error_writable_details()
    -   gst_message_parse_warning_writable_details()
    -   gst_message_parse_info_writable_details() This is used in uridecodebin3 to add the triggering URI to any INFO, WARNING
        or ERROR messages posted on the bus, and in decodebin3 to add the stream ID to any missing plugin messages posted on the
        bus.

-   gst_util_floor_log2() returns smallest integral value not bigger than log2(v).

-   gst_util_fraction_multiply_int64() is a 64-bit variant of gst_util_fraction_multiply().

#### GstIdStr replaces GQuark in structure and caps APIs

-   GQuarks are integer identifiers for strings that are inserted into a global hash table, allowing in theory for cheap
    equality comparisons. In GStreamer they have been used to represent GstStructure names and field names. The problem is that
    these strings once added to the global string table can never be freed again, which can lead to ever-increasing memory usage
    for processes where such name identifiers are created based on external input or on locally-created random identifiers.

-   GstIdStr is a new data structure made to replace quarks in our APIs. It can hold a short string inline, a static string, or
    a reference to a heap-allocated longer string, and allows for cheap storage of short strings and cheap comparisons. It does
    not involve access to a global hash table protected by a global lock, and as most strings used in GStreamer structures are
    very short, it is actually more performant than quarks in many scenarios.

-   GQuark-using APIs in GstStructure or GstCaps have been deprecated and equivalent APIs using GstIdStr have been added
    instead. For more details about this change watch Sebastian’s GStreamer Conference presentation “GQuark in GStreamer
    structures - what nonsense!”.

-   Most applications and plugins will have been using the plain string-based APIs which are not affected by this change.

#### GstVecDeque

-   Moved GstQueueArray as GstVecDeque into core for use in GstBus, the ringbuffer logger and in GstBufferPool, where an overly
    complicated signaling mechanism using GstAtomicQueue and GstPoll was replaced with GstVecDeque and a simple mutex/cond.

-   GstQueueArray in libgstbase was deprecated in favour of GstVecDeque.

-   GstAtomicQueue will be deprecated once all users in GStreamer have been moved over to GstVecDeque.

### Audio Library

-   Added gst_audio_reorder_channels_with_reorder_map() which allows reordering the samples with a pre-calculated reorder map
    instead of re-calculating the reorder map every time.

-   Add top-surround-left and top-surround-right channel positions

-   GstAudioConverter now supports more numerical types for the mix matrix, namely double, int, int64, uint, and uint64 in
    addition to plain floats.

### Plugins Base Utils Library

-   New AV1 caps utility functions for AV1 Codec Configuration Record codec_data handling

-   The GstEncodingProfile (de)serialization functions are now public

-   GstEncodingProfile gained a way to specify a factory-name when specifying caps. In some cases you want to ensure that a
    specific element factory is used while requiring some specific caps, but this was not possible so far. You can now do
    e.g. qtmux:video/x-prores,variant=standard|factory-name=avenc_prores_ks to ensure that the avenc_prores_ks factory is used
    to produce the variant of prores video stream.

### Tag Library

-   EXIF handling now support the CAPTURING_LIGHT_SOURCE tag

-   Vorbis tag handling gained support for the LYRICS tag

### Video Library and OpenGL Library

-   gst_video_convert_sample(), gst_video_convert_sample_async() gained support for D3D12 conversion.

-   GstVideoEncoder: gst_video_encoder_release_frame() and gst_video_encoder_drop_frame() have been made available as public
    API.

-   Navigation: gained mouse double click event support

-   Video element QoS handling was improved so as to not overshoot the QoS earliest time by a factor of 2. This was fixed in the
    video decoder, encoder, aggregator and audiovisualizer base classes, as well as in the adaptivedemux, deinterlace,
    monoscope, shapewipe, and (old) videomixer elements.

-   GstVideoConverter gained fast paths for v210 to/from I420_10 / I422_10

-   New gst_video_dma_drm_format_from_gst_format() helper function that converts a video format into a dma drm fourcc / modifier
    pair, plus gst_video_dma_drm_format_to_gst_format() which will do the reverse.

-   In the same vein gst_gl_dma_buf_transform_gst_formats_to_drm_formats() and
    gst_gl_dma_buf_transform_drm_formats_to_gst_formats() have been added to the GStreamer OpenGL support library.

-   GLDisplay/EGL: Add API (gst_gl_display_egl_set_foreign()) for overriding foreign-ness of the EGLDisplay in order to control
    whether GStreamer should call eglTerminate() or not.

-   Additional DMA DRM format definitions/mappings:

    -   NV15, NV20, NV30
    -   NV12_16L32S, MT2110T, MT2110R as used on Meditek SoCs
    -   NV12_10LE40
    -   RGB15, GRAY8, GRAY16_LE, GRAY16_BE
    -   plus support for big endian DRM formats and DRM vendor modifiers

New Raw Video Formats

-   Packed 4:2:2 YUV with 16 bits per channel:
    -   Y216_LE, Y216_BE
-   Packed 4:4:4:4 YUV with alpha, with 16 bits per channel:
    -   Y416_LE, Y416_BE
-   10-bit grayscale, packed into 16-bit words with left padding:
    -   GRAY10_LE16

### GstPlay Library

-   Add stream-id based selection of streams to better match playbin3’s API:
    -   Add accessors for the stream ID and selection API based on the stream ID:
        -   gst_play_stream_info_get_stream_id()
        -   gst_play_set_audio_track_id()
        -   gst_play_set_video_track_id()
        -   gst_play_set_subtitle_track_id()
        -   gst_play_set_track_ids()
    -   Deprecate the old index-based APIs:
        -   gst_play_stream_info_get_index()
        -   gst_play_set_audio_track()
        -   gst_play_set_video_track()
        -   gst_play_set_subtitle_track()
    -   Remove old playbin support
    -   Implement the track enable API based on stream selection
-   Distinguish missing plugin errors and include more details (uri, and stream-id if available) in error/warning messages:
    -   gst_play_message_get_uri()
    -   gst_play_message_get_stream_id()
    -   GST_PLAY_ERROR_MISSING_PLUGIN
    -   gst_play_message_parse_error_missing_plugin()
    -   gst_play_message_parse_warning_missing_plugin()
-   Improve play message API inconsistencies:
    -   Consistently name parse functions according to their message type:
        -   gst_play_message_parse_duration_changed()
        -   gst_play_message_parse_buffering()
    -   Deprecate the misnamed functions:
        -   gst_play_message_parse_duration_updated()
        -   gst_play_message_parse_buffering_percent()
    -   Add missing parse functions:
        -   gst_play_message_parse_uri_loaded()
        -   gst_play_message_parse_seek_done()
-   Support disabling the selected track at startup

## Miscellaneous performance, latency and memory optimisations

-   dvdspu: use multiple minimal sized PGS overlay rectangles instead of a single large one to minimise the total blitting
    surface in case of disjointed rectangles.

-   video-frame: reduce number of memcpy() calls on frame copies if possible

-   video-converter: added fast path conversions between v210 and I420_10 / I422_10

-   As always there have been plenty of performance, latency and memory optimisations all over the place.

## Miscellaneous other changes and enhancements

-   netclientclock: now also emits the clock synced signal when corrupted to signal that sync has been lost.

-   GstValue, GstStructure: can now (de)serialize string arrays (G_TYPE_STRV)

## Tracing framework and debugging improvements

-   dot files (pipeline graph dumps) are now written to disk atomically

-   tracing: add hooks for gst_pad_send_event_unchecked() and GstMemory init/free

-   tracers: Simplify params handling using GstStructure and object properties and move tracers over to property-based
    configuration (leaks, latency).

-   textoverlay, clockoverlay, timeoverlay: new "response-time-compensation" property that makes the element render the text or
    timestamp twice: Once in the normal location and once in a different sequentially predictable location for every frame. This
    is useful when measuring video latency by taking a photo with a camera of two screens showing the test video overlayed with
    timeoverlay or clockoverlay. In these cases, you will often see ghosting if the display’s pixel response time is not great,
    which makes it difficult to discern what the current timestamp being shown is. Rendering in a different location for each
    frame makes it easy to tell what the latest timestamp is. In addition, you can also use the fade-time of the previous frame
    to measure with sub-framerate accuracy when the photo was taken, not just clamped to the framerate, giving you a higher
    precision latency value.

New tracers

-   memory-tracer: New tracer that can track memory usage over time

-   pad-push-timings: New tracer for tracing pad push timings

-   pcap-writer: New tracer that can store the buffers flowing through a pad as PCAP file

Dot tracer/viewer

-   New dots tracer that simplifies the pipeline visualization workflow:
    -   Automatically configures dot file directory and cleanup
    -   Integrates with the pipeline-snapshotS tracer to allow dumping pipeline on demand from the gst-dots-viewer web interface
    -   Uses GST_DEBUG_DUMP_DOT_DIR or falls back to $XDG_CACHE_HOME/gstreamer-dots
-   New gst-dots-viewer web tool for real-time pipeline visualization
    -   Provides interface to browse and visualize pipeline dot files
    -   Features on-demand pipeline snapshots via “Dump Pipelines” button
    -   WebSocket integration for live updates
    -   Uses GST_DEBUG_DUMP_DOT_DIR or falls back to $XDG_CACHE_HOME/gstreamer-dots
-   Simple usage:
    -   gst-dots-viewer (starts server)
    -   GST_TRACERS=dots gst-launch-1.0 videotestsrc ! autovideosink (runs with tracer)
    -   View at http://localhost:3000

Debug logging system improvements

-   Nothing major in this cycle.

## Tools

-   gst-inspect-1.0 documents tracer properties now and shows element flags

-   gst-launch-1.0 will show error messages posted during pipeline construction

## GStreamer FFmpeg wrapper

-   Add support for H.266/VVC decoder

-   Add mappings for the Hap video codec, the Quite OK Image codec (QOI) and the M101 Matrox uncompressed SD video codec.

-   Don’t register elements for which we have no caps and which were non-functional as a result (showing unknown/unknown caps).

-   The S302M audio encoder now supports up to 8 channels.

-   Various tag handling improvements in the avdemux wrappers, especially when there are both upstream tags and additional local
    tags.

-   Support for 10-bit grayscale formats

## GStreamer RTSP server

-   GstRTSPOnvifMediaFactoryClass gained a ::create_backchannel_stream() vfunc. This allows subclasses to delay creation of the
    backchannel to later in the sequence, which is useful in scenarios where the RTSP server acts as a relay and the supported
    caps depend on the upstream camera, for example.

-   The ONVIF backchannel example now features support for different codecs, including AAC.

## VA-API integration

VA plugin

-   New VA elements:

    -   H.266 / VVC video decoder
    -   VP8 video encoder
    -   JPEG encoder
    -   VP9 + VP8 alpha decodebin

    Remember that the availability of these elements depends on your platform and driver.

-   There are a lot of improvements and bug fixes, to hightlight some of them:

    -   Improved B pyramid mode for both H264 and HEVC encoding when reference frame count exceeds 2, optimizing pyramid level
        handling.
    -   Enabled ICQ and QVBR modes for several encoders, including H264, H265, VP9 and AV1.
    -   Updated rate control features by setting the quality factor parameter, while improving bitrate change handling.
    -   Improved VP9 encoder’s ability to avoid reopening or renegotiating encoder settings when parameters remain stable.
    -   Added functionality to adjust the trellis parameter in encoders.
    -   Optimize encoders throughput with the introduction of output delay.
    -   Added support for new interpolation methods for scaling and improvements for handling interlace modes.

GStreamer-VAAPI is now deprecated

-   gstreamer-vaapi has been deprecated and is no longer actively maintained. Users who rely on gstreamer-vaapi are encouraged
    to migrate to the va plugin’s elements at the earliest opportunity.

-   vaapi*enc encoders have been demoted to a rank of None, so will no longer be autoplugged in encodebin. They have also no
    longer advertise dmabuf caps or unusual pixel formats on their input pad template caps.

## GStreamer Video4Linux2 support

-   Implemented DMA_DRM caps negotiation

-   Framerate negotiation improvements

-   Support for P010 and BGR10A2_LE pixel formats

-   The V4L2 stateless decoders now support inter-frame resolution changes for AV1 and VP9

-   The V4L2 stateful encoders can now handle dynamic frame rates (0/1), and colorimetry negotiation was also improved.

## GStreamer Editing Services and NLE

-   Added support for reverse playback with a new reverse property on nlesource which is exposed child property on GESClip

-   Input channels reordering for flexible audio channel mapping

-   Added support for transition in the ges-launch-1.0 timeline description format

-   Added support for GstContext sharing in GESTimeline

-   Added basic support for duplicated children property in GESTimelineElement

-   validate: Add an action type to group clips

## GStreamer validate

-   Added new action types:

    -   start-http-server: Start a new instance of our HTTP test server
    -   http-requests: Send an HTTP request to a server, designed to work with our test http server

-   HTTP server control endpoints to allow scenarios to control the server behavior, allowing simulating server failures from
    tests

-   Improved the select-streams action type, adding support for selecting the same streams several times

-   Added support for forcing monitoring of all pipelines in validatetest files

-   Enhanced support for expected Error messages on the bus

-   Added ways to retrieve HTTP server port in .validatetest files

-   Added support for lldb in the gst-validate-launcher

## GStreamer Python Bindings

gst-python is an extension of the regular GStreamer Python bindings based on gobject-introspection information and PyGObject,
and provides “syntactic sugar” in form of overrides for various GStreamer APIs that makes them easier to use in Python and more
pythonic; as well as support for APIs that aren’t available through the regular gobject-introspection based bindings, such as
e.g. GStreamer’s fundamental GLib types such as Gst.Fraction, Gst.IntRange etc.

-   The python Meson build option has been renamed to python-exe (and will yield to the monorepo build option of the same name
    if set, in a monorepo build context).

-   Added an iterator for AnalyticsRelationMeta

-   Implement __eq__ for Mtd classes

-   Various build fixes and Windows-related fixes.

## GStreamer C# Bindings

-   The C# bindings have been updated for the latest GStreamer 1.26 API

## GStreamer Rust Bindings and Rust Plugins

The GStreamer Rust bindings and plugins are released separately with a different release cadence that’s tied to the gtk-rs
release cycle.

The latest release of the bindings (0.23) has already been updated for the new GStreamer 1.26 APIs, and works with any GStreamer
version starting at 1.14.

gst-plugins-rs, the module containing GStreamer plugins written in Rust, has also seen lots of activity with many new elements
and plugins. The GStreamer 1.26 binaries will be tracking the main branch of gst-plugins-rs for starters and then track the 0.14
branch once that has been released (around summer 2025). After that, fixes from newer versions will be backported as needed to
the 0.14 branch for future 1.26.x bugfix releases.

Rust plugins can be used from any programming language. To applications they look just like a plugin written in C or C++.

### New Rust elements

-   awstranscriber2, awstranslate: New elements around the AWS transcription and translation services.

-   cea708mux: New element that allows to mux multiple CEA708 services into a single stream.

-   cdpserviceinject: New element for injecting a CDP service descriptor into closed caption CDP packets

-   cea708overlay: New element for overlaying CEA608 / CEA708 closed captions over video streams.

-   gopbuffer: New element that can buffer a minimum duration of data delimited by discrete GOPs (Group of Picture)

-   hlscmafsink, hlssink3: New single-variant HLS sink elements that can output CMAF (fMP4) or MPEG-TS fragments.

-   hlsmultivariantsink: New sink element that can output an HLS stream with multiple variants

-   mpegtslivesrc: New source element that can wrap a live MPEG-TS source (e.g. SRT or UDP source) and provides a clock based on
    the PCR.

-   onvifmetadataextractor: New element that can extract ONVIF metadata from GstMetas into a separate stream

-   originalbuffer: New plugin with originalbuffersave / originalbufferrestore elements that allow saving an original buffer,
    modifying it for analytics, and then restoring the original buffer content while keeping any additional metadata that was
    added.

-   polly: New element around the AWS text-to-speech polly services

-   quinn: New plugin that contains various QUIC-based elements for working with raw QUIC streams, RTP-over-QUIC (RoQ) and
    WebTransport.

-   relationmeta: New plugin with elements converting between GstRelationMeta and ONVIF XML metadata.

-   New Rust RTP payloaders and depayloaders for AC3, AMR, JPEG, KLV, MPEG-TS (MP2T), MPEG-4 (MP4A, MP4G), Opus, PCMU (uLaw),
    PCMA (aLaw), VP8, VP9.

-   New rtpbin2 based on rtprecv / rtpsend elements

-   speechmatics: New transcriber / speech-to-text and translation element

-   New spotifylyricssrc element for grabbing lyrics from Spotify.

-   streamgrouper: New element that takes any number of streams as input and adjusts their stream-start events in such a way
    that they all belong to the same stream group.

-   translationbin: Helper bin around translation elements, similar to the already existing transcriberbin for transcriptions.

-   tttocea708: New element for converting timed-text to CEA708 closed captions

-   A VVdeC-based H.266 decoder element was added to the Rust plugins, based on the Fraunhofer Versatile Video Decoder library.

For a full list of changes in the Rust plugins see the gst-plugins-rs ChangeLog between versions 0.12 (shipped with GStreamer
1.24) and 0.14.x (shipped with GStreamer 1.26).

Note that at the time of GStreamer 1.26.0 gst-plugins-rs 0.14 was not released yet and the git main branch was included instead
(see above). As such, the ChangeLog also did not contain the changes between the latest 0.13 release and 0.14.0.

## Build and Dependencies

-   Meson >= 1.4 is now required for all modules

-   liborc >= 0.4.41 is strongly recommended

-   libnice >= 0.1.22 is strongly recommended, as it is required for WebRTC ICE consent freshness (RFC 7675).

-   The ASIO plugin dropped its external SDK header dependency, so it can always be built and shipped more easily.

-   Require tinyalsa >= 1.1.0 when building the tinyalsa plugin

-   The srtp plugin now requires libsrtp2, support for libsrtp1 was dropped.

Monorepo build

-   The FFmpeg subproject wrap was updated to 7.1

-   Many other wrap updates

gstreamer-full

-   No major changes

Development environment

-   Local pre-commit checks via git hooks have been moved over to pre-commit, including the code indentation check.

-   Code indentation checking no longer relies on a locally installed copy of GNU indent (which had different outcomes depending
    on the exact version installed). Instead pre-commit will automatically install the gst-indent-1.0 indentation tool through
    pip, which also works on Windows and macOS.

-   A pre-commit hook has been added to check documentation cache updates and since tags.

-   Many meson wrap updates, including to FFmpeg 7.1

-   The uninstalled development environment should work better on macOS now, also in combination with homebrew (e.g. when
    libsoup comes from homebrew).

-   New python-exe Meson build option to override the target Python installation to use. This will be picked up by the
    gst-python and gst-editing-sevices subprojects.

## Platform-specific changes and improvements

### Android

-   The recommended mechanism to build Android apps has changed from Android.mk to CMake-in-Gradle using
    FindGStreamerMobile.cmake. Android.mk support has been deprecated and will be removed in the next stable release. For more
    information, see below, in the Cerbero section.
-   More H.264/H.265 profiles and levels have been added to the androidmedia hardware-accelerated video encoder and decoder
    elements, along with mappings for a number of additional pixel formats for P010, packed 4:2:0 variants and RGBA layouts,
    which fixes problems with android decoders refusing to output raw video frames with decoders that announce support for these
    common pixel formats and only allow the ‘hardware surfaces output’ path.

### Apple macOS and iOS

-   atenc: added an Apple AAC audio encoder

-   atdec can now decode audio with more than two channels

-   vtenc has received various bug fixes as well as a number of new features:

    -   Support for HEVC with alpha encoding via the new vtenc_h265a element
    -   additional rate control options for constant bitrate encoding (only supported on macOS 13.0+ and iOS 16.0+ on Apple
        Silicon), setting data rate limits, and emulating CBR mode via data rate limits where CBR is not supported.
    -   HLG color transfer support
    -   new "max-frame-delay" property (for ProRes)

-   Better macOS support for gst-validate tools which now use gst_macos_main() and support lldb

-   The osxaudio device provider exposes more properties including a unique id

-   osxaudio will automatically set up AVAudioSession on iOS and always expose the maximum number of channels a device supports
    with an unpositioned layout.

-   The monorepo development environment should work better on macOS now

-   CMake apps that build macOS and iOS apps can consume GStreamer more easily now, using FindGStreamer.cmake or
    FindGStreamerMobile.cmake respectively.

-   In the near future, CMake in Xcode will be the preferred way of building the iOS tutorials. See below, in the Cerbero
    section.

### Windows

-   webview2src: new Microsoft WebView2 based web browser source element

-   The mediafoundation plugin can also be built with MinGW now.

-   The GTK3 plugin has gained support for OpenGL/WGL on Windows

-   qsv: Add support for d3d12 interop in encoder, via D3D11 textures

### Cerbero

Cerbero is a meta build system used to build GStreamer plus dependencies on platforms where dependencies are not readily
available, such as Windows, Android, iOS, and macOS.

General improvements

-   New features:

    -   Python bindings support has been re-introduced and now supports Linux, Windows (MSVC) and macOS. Simply downloading the
        official binaries and setting PYTHONPATH to the appropriate directory is sufficient.

        -   This should make it easier for macOS and Windows users to use Python libraries, frameworks, and projects that use
            GStreamer such as Pitivi and gst-python-ml.

    -   Introspection support has been re-introduced on Linux, Windows (MSVC), and macOS.

    -   New variants assert and checks to disable GLib assertions and runtime checks for performance reasons. Please note that
        these are not recommended because they have significant behavioural side-effects, make debugging harder, and should only
        be enabled when targeting extremely resource-constrained platforms.

-   API/ABI changes:

    -   Libsoup has been upgraded from 2.74 to 3.6, which is an API and ABI breaking change. The soup and adaptivedemux2 plugins
        are unchanged, but your applications may need updating since libsoup-2.4 and libsoup-3.0 cannot co-exist in the same
        process.

    -   OpenSSL has been updated from 1.1.1 to 3.4, which is an ABI and API breaking change. Plugins are unchanged, but your
        applications may need updating.

-   Plugins added:

    -   The svt-av1 plugin is now shipped in the binary releases for all platforms.

    -   The svt-jpeg-xs plugin is now shipped in the binary releases for all platforms.

    -   The x265 plugin is now shipped in the binary releases for all platforms.

    -   All gst-plugins-rs elements are now shipped in the binary releases for all platforms, except those that have C/C++
        system-deps like GTK4. For a full list, see the Rust section above.

-   Plugins changed:

    -   The rsvg plugin now uses librsvg written in Rust. The only side-effects of this should be better SVG rendering and
        slightly larger plugin size.

    -   The webrtc Rust plugin now also supports aws and livekit integrations .

-   Plugins removed:

    -   webrtc-audio-processing has been updated to 2.0, which means the isac plugin is no longer shipped.

-   Development improvements:

    -   Support for the shell command has been added to cross-macos-universal, since the prefix is executable despite being a
        cross-compile target

    -   More recipes have been ported away from Autotools to Meson and CMake, speeding up the build and increasing platform
        support.

#### macOS

-   Python bindings support on macOS only supports using the Xcode-provided Python 3

-   MoltenVK support in the applemedia plugin now also works on arm64 when doing a cross-universal build.

#### iOS

-   CMake inside Xcode will soon be the recommended way to consume GStreamer when building iOS apps, similar to Android apps.

    -   FindGStreamerMobile.cmake is the recommended way to consume GStreamer now

    -   Tutorials and examples still use Xcode project files, but CMake support will be the active focus going forward

#### Windows

-   The minimum supported OS version is now Windows 10.

    -   GStreamer itself can still be built for an older Windows, so if your project is majorly impacted by this, please open an
        issue with details.

-   The Windows MSI installers are now based on WiX v5, with several improvements including a much faster MSI creation process,
    improved naming in Add/Remove Programs, and more.

    -   Windows installer packages: Starting with 1.26, due to security reasons, the default installation directory has changed
        from C:\gstreamer to the Program Files folder, e.g. C:\Program Files (x86)\gstreamer for the 32-bit package on 64-bit
        Windows. If you upgrade from 1.24 or older versions, the 1.26 installers will NOT keep using the existing folder.
        Nevertheless if you were using C:\gstreamer we strongly recommend you double-check the install location.

    -   Note for MSI packagers: Starting with 1.26, the installers were ported to WiX 5.0. As part of this, the property for
        setting the installation directory has been changed to INSTALLDIR, and it now requires a full path to the desired
        directory, e.g. C:\gstreamer instead of C:\.

    -   Cross-MinGW build no longer supports the creation of MSIs. Please use tarballs.

-   MinGW:

    -   MinGW toolchain has been updated from GCC 8.2 → 14.2 and MinGW 6.0 → 12.0

    -   The mediafoundation plugin is also shipped in the MinGW packages now.

    -   The d3d12 plugin is also shipped in the MinGW packages now.

    -   Rust support has been enabled on MinGW 64-bit. Rust support cannot work on 32-bit MinGW due to differences in exception
        handling between our 32-bit MinGW toolchain and that used by the Rust project

-   The asio plugin is shipped now, since it no longer has a build-time dependency on the ASIO SDK.

-   The new plugin webview2 is shipped with MSVC. It requires the relevant component shipped with Windows.

#### Linux

-   Preliminary support for Alma Linux has been added.

-   RHEL distro support has been improved.

-   Cerbero CI now tests the build on Ubuntu 24.04 LTS.

-   curl is used for downloading sources on Fedora instead of wget, since they have moved to wget2 despite show-stopper
    regressions such as returning a success error code on download failure.

#### Android

-   CMake inside Gradle is now the recommended way to consume GStreamer when building Android apps

    -   FindGStreamerMobile.cmake is the recommended way to consume GStreamer now

    -   1.26 will support both CMake and Make inside Gradle, but the Make support will likely be removed in 1.28

    -   Documentation updates are still a work-in-progress, help is appreciated

-   Android tutorials and examples are now built with gradle + cmake instead of gradle + make on the CI

## Documentation improvements

-   Tracer objects information is now included in the documentation

## Possibly Breaking Changes

-   qroverlay: the "pixel-size" property has been removed in favour of a new "size" property with slightly different semantics,
    where the size of the square is expressed in percent of the smallest of width and height.

-   svtav1enc: The SVT-AV1 3.0.0 API exposes a different mechanism to configure the level of parallelism when encoding, which
    has been exposed as a new "level-of-parallelism" property. The old "logical-processors" property is no longer functional if
    the plugin has been compiled against the new API, which might affect encoder performance if application code setting it is
    not updated.

-   udpsrc: now disables allocated port reuse for unicast to avoid unexpected side-effects of SO_REUSEADDR where the kernel
    allocates the same listening port for multiple udpsrc.

-   uridecodebin3 remove non-functional "source" property that doesn’t make sense and always returned NULL anyway.

## Known Issues

-   GstBuffer now uses C11 atomics for 64 bit atomic operations if available, which may require linking to libatomic on some
    systems, but this is not done automatically yet, see issue #4177.

Contributors

Aaron Boxer, Adrian Perez de Castro, Adrien De Coninck, Alan Coopersmith, Albert Sjolund, Alexander Slobodeniuk, Alex Ashley,
Alicia Boya García, Andoni Morales Alastruey, Andreas Wittmann, Andrew Yooeun Chun, Angelo Verlain, Aniket Hande, Antonio
Larrosa, Antonio Morales, Armin Begovic, Arnaud Vrac, Artem Martus, Arun Raghavan, Benjamin Gaignard, Benjamin Gräf, Bill
Nottingham, Brad Hards, Brad Reitmeyer, Branko Subasic, Carlo Caione, Carlos Bentzen, Carlos Falgueras García, cdelguercio, Chao
Guo, Cheah, Cheung Yik Pang, chitao1234, Chris Bainbridge, Chris Spencer, Chris Spoelstra, Christian Meissl, Christopher Degawa,
Chun-wei Fan, Colin Kinloch, Corentin Damman, Daniel Morin, Daniel Pendse, Daniel Stone, Dan Yeaw, Dave Lucia, David Rosca, Dean
Zhang (张安迪), Denis Yuji Shimizu, Detlev Casanova, Devon Sookhoo, Diego Nieto, Dongyun Seo, dukesook, Edward Hervey, eipachte,
Eli Mallon, Elizabeth Figura, Elliot Chen, Emil Ljungdahl, Emil Pettersson, eri, F. Duncanh, Fotis Xenakis, Francisco Javier
Velázquez-García, Francis Quiers, François Laignel, George Hazan, Glyn Davies, Guillaume Desmottes, Guillermo E. Martinez,
Haihua Hu, Håvard Graff, He Junyan, Hosang Lee, Hou Qi, Hugues Fruchet, Hyunwoo, iodar, jadarve, Jakub Adam, Jakub Vaněk, James
Cowgill, James Oliver, Jan Alexander Steffens (heftig), Jan Schmidt, Jeffery Wilson, Jendrik Weise, Jerome Colle, Jesper Jensen,
Jimmy Ohn, Jochen Henneberg, Johan Sternerup, Jonas K Danielsson, Jonas Rebmann, Jordan Petridis, Jordan Petridіs, Jordan
Yelloz, Jorge Zapata, Joshua Breeden, Julian Bouzas, Jurijs Satcs, Kévin Commaille, Kevin Wang, Khem Raj, kingosticks, Leonardo
Salvatore, L. E. Segovia, Liam, Lim, Loïc Le Page, Loïc Yhuel, Lyra McMillan, Maksym Khomenko, Marc-André Lureau, Marek Olejnik,
Marek Vasut, Marianna Smidth Buschle, Marijn Suijten, Mark-André Schadow, Mark Nauwelaerts, Markus Ebner, Martin Nordholts, Mart
Raudsepp, Mathieu Duponchelle, Matthew Waters, Maxim P. DEMENTYEV, Max Romanov, Mengkejiergeli Ba, Michael Grzeschik, Michael
Olbrich, Michael Scherle, Michael Tretter, Michiel Westerbeek, Mikhail Rudenko, Nacho Garcia, Nick Steel, Nicolas Dufresne,
Niklas Jang, Nirbheek Chauhan, Ognyan Tonchev, Olivier Crête, Oskar Fiedot, Pablo García, Pablo Sun, Patricia Muscalu, Paweł
Kotiuk, Peter Kjellerstedt, Peter Stensson, pgarciasancho, Philippe Normand, Philipp Zabel, Piotr Brzeziński, Qian Hu (胡骞),
Rafael Caricio, Randy Li (ayaka), Rares Branici, Ray Tiley, Robert Ayrapetyan, Robert Guziolowski, Robert Mader, Roberto Viola,
Robert Rosengren, RSWilli,Ruben González, Ruijing Dong, Sachin Gadag, Sam James, Samuel Thibault, Sanchayan Maity, Scott Moreau,
Sebastian Dröge, Sebastian Gross, Sebastien Cote, Sergey Krivohatskiy, Sergey Radionov, Seungha Yang, Seungmin Kim, Shengqi Yu,
Sid Sethupathi, Silvio Lazzeretti, Simonas Kazlauskas, Stefan Riedmüller, Stéphane Cerveau, Tamas Levai, Taruntej Kanakamalla,
Théo Maillart, Thibault Saunier, Thomas Goodwin, Thomas Klausner, Tihran Katolikian, Tim Blechmann, Tim-Philipp Müller, Tjitte
de Wert, Tomas Granath, Tomáš Polomský, tomaszmi, Tom Schuring, U. Artie Eoff, valadaptive, Víctor Manuel Jáquez Leal, Vivia
Nikolaidou, W. Bartel, Weijian Pan, William Wedler, Will Miller, Wim Taymans, Wojciech Kapsa, Xavier Claessens, Xi Ruoyao,
Xizhen, Yaakov Selkowitz, Yacine Bandou, Zach van Rijn, Zeno Endemann, Zhao, Zhong Hongcheng,

… and many others who have contributed bug reports, translations, sent suggestions or helped testing. Thank you all!

Stable 1.26 branch

After the 1.26.0 release there will be several 1.26.x bug-fix releases which will contain bug fixes which have been deemed
suitable for a stable branch, but no new features or intrusive changes will be added to a bug-fix release usually. The 1.26.x
bug-fix releases will be made from the git 1.26 branch, which is a stable release series branch.

Schedule for 1.28

Our next major feature release will be 1.28, and 1.27 will be the unstable development version leading up to the stable 1.28
release. The development of 1.27/1.28 will happen in the git main branch of the GStreamer mono repository.

The schedule for 1.28 is yet to be decided.

1.28 will be backwards-compatible to the stable 1.26, 1.24, 1.22, 1.20, 1.18, 1.16, 1.14, 1.12, 1.10, 1.8, 1.6, 1.4, 1.2 and 1.0
release series.

--------------------------------------------------------------------------------------------------------------------------------

These release notes have been prepared by Tim-Philipp Müller with contributions from Arun Raghavan, Daniel Morin, Nirbheek
Chauhan, Olivier Crête, Philippe Normand, Sebastian Dröge, Seungha Yang, Thibault Saunier, and Víctor Manuel Jáquez Leal.

License: CC BY-SA 4.0