Follow developments in AV1, AV2, H.264/AVC, HEVC/H.265, VVC/H.266, FFmpeg, cloud transcoding, hardware acceleration, bitrate optimization, and video quality measurement. StreamingMeme curates and summarizes the updates that matter to streaming engineers, product teams, and media technology leaders.
The Alliance for Open Media has released the Open Audio Renderer (OAR) v1, a reference implementation for rendering spatial audio based on the IAMF specification. This release provides developers with an interoperable baseline for spatial audio across diverse playback hardware, including loudspeakers and headphones.
Nokia has released public Neural Network Post Filtering (NNPF) software to improve VVC video quality through AI-based post-processing. The tool enables streaming platforms to restore detail and texture during final decoding using standardized metadata and inference on consumer hardware.
NVIDIA released nvmath-python v1.0, a new library that bridges Python scientific workflows with its CUDA-X math libraries. The tool enables high-performance matrix and FFT operations across CPU, single-GPU, and multi-node GPU environments.
Ateme discusses the ongoing industry move toward multi-codec streaming, noting that 40% of survey respondents plan to deploy AV1 by 2026. The company advocates for just-in-time packaging as a strategy to mitigate the increased storage requirements of managing legacy H.264 alongside newer formats like HEVC and VVC.
Intel developer Daniil Seliverstov submitted patches for the Linux Xe kernel driver to enable HuC micro-controller support for DG2/Alchemist GPUs. This update resolves a compatibility limitation for hardware-accelerated transcoding of codecs like HEVC on modern Intel graphics hardware.
Thimeo Software has released Stereo Tool v11.00, which introduces MicroMPX version 5 and features improved encoding efficiency, lower CPU usage, and broader support for RTP and RIST protocols. The update also includes new I/O framework capabilities, an R128 limiter, and the RuleBreaker 2 clipper designed for broadcast and streaming audio workflows.
Beamr details its 'ML-Safe' content-adaptive compression framework for processing large-scale autonomous vehicle video data. The company claims the method maintains model accuracy within 2% of raw baselines while significantly reducing storage and transfer requirements.
Adobe released six updates to Frame.io, including an AI-powered search tool, a Comparison Viewer with pixel-difference detection, and a redesigned iOS app. The update also adds new C2C integrations with Leica, Accsoon, and Pomfort, as well as tighter integration across the Adobe creative suite.
Google's Project Zero team disclosed a zero-click exploit chain on Pixel 10 devices that leveraged vulnerabilities in a Dolby audio decoder and a custom hardware video processing unit (VPU) driver. The report provides a technical analysis of how the circumvention of standard video decoding protocols like V4L2 and repeated coding errors in hardware drivers created critical security gaps that were subsequently patched.
The ITU, ISO, and IEC have issued a joint call for proposals for a next-generation video-coding standard scheduled for development by 2029. The initiative aims to surpass the performance of the current VVC standard, with initial evaluation of proposed compression technologies set for January 2027.
Chips&Media has launched the WAVE-J JPEG codec IP, designed for 8K60 imaging systems with support for resolutions up to 64K x 64K. The core features a 128-bit AXI4 interface and integrated image processing, targeting high-performance applications in edge devices, AI servers, and broadcast infrastructure.
The Mesa 3D Graphics Library has released version 26.1.6, introducing support for Vulkan 1.4 and OpenGL 4.6. This bug-fix update specifically resolves regressions in H.265 hardware decoding on Raspberry Pi 5 and addresses encoding issues related to AV1 and HEVC on AMD and Intel hardware.
Wowza has introduced its Video Intelligence Framework (VIF), which enables real-time computer vision analysis directly within the Wowza Streaming Engine pipeline. The framework supports multiple output channels, including metadata and webhooks, allowing operators to trigger automated workflows without relying on external post-event analysis.
NETINT Technologies has released a seven-step technical framework for migrating streaming video workloads from CPU-based processing to their Quadra Video Processing Units (VPUs). The guide provides engineering teams with recommended strategies for validating hardware, ensuring service equivalence through feature parity, and establishing reliable rollback paths during the transition.
The ITU, ISO, and IEC have issued an official call for proposals for a next-generation video compression standard intended to succeed VVC. The initiative aims to support emerging applications like XR and AI-driven content processing by 2029.
HDR10+ Technologies clarifies that its specification natively supports up to 16-bit color depth, attributing the current 10-bit delivery limitation to constraints inherent in modern streaming video codecs. The piece further distinguishes between bit-depth capacity and peak brightness parameters, noting their independence within the SMPTE ST 2084 transfer function.
A regression in MediaMTX v1.16.0 causes AV1 video ingested over RTMP to become undecodable when delivered via RTSP, while HLS delivery remains unaffected. The issue, linked to code changes in the internal codec processor, has been verified across multiple subsequent versions using FFmpeg and other publishers.
FFmpeg has released patches for six high-severity vulnerabilities in versions up to 8.1.2, including heap buffer overflows and denial-of-service flaws with CVSS scores reaching 8.7. These vulnerabilities pose significant supply-chain risks for streaming platforms, CDNs, and media services that utilize the framework for automated transcoding and media processing.
Brazil has officially launched its TV 3.0 (DTV+) standard, which integrates 4K VVC video encoding, immersive MPEG-H audio, and internet-based interactivity for free-to-air broadcasting. The rollout for this next-generation terrestrial television system is planned in phases, with national coverage targeted by 2030, necessitating new hardware converters for existing television sets.
Researchers have proposed GenSplatCodec, a unified dual-stream codec that reformulates 3D Gaussian Splatting compression as geometry-guided generative decoding. The system demonstrates a 17x reduction in storage costs compared to existing feed-forward methods while maintaining rendering fidelity for immersive media applications.
AMD has announced its upcoming Venice data center processor platform, which will feature a 2nm manufacturing process and Zen 6 architecture. Expected to begin rolling out in 2026, the chips aim to improve power efficiency and performance density to compete with Intel and custom Arm-based silicon.
Microsoft has open sourced MLVC, a neural network-based video codec designed to leverage Neural Processing Units (NPUs) for efficient video compression. The codec demonstrates significant bandwidth reductions compared to H.264 and is currently being deployed within Microsoft Teams.
Phoronix has published benchmarks for the Framework Laptop 13 Pro, featuring the Intel Core Ultra X9 388H SoC. The testing evaluates performance using standard open-source video encoding codecs, specifically SVT-AV1, Kvazaar, and VVenC, to assess efficiency for transcoding workloads.
Developer chanyavrc has released vp9dec, a clean-room, zero-dependency VP9 video decoder written in Rust. The project is fully compliant with all four VP9 profiles and has validated bit-exact conformance against 315 official test vectors.
TanStack has released alpha versions of two new libraries, TanStack Markdown and TanStack Highlight, which aim to reduce bundle sizes for syntax highlighting and markdown parsing. These tools are designed for technical documentation and AI-streamed text rendering with significantly lower dependency requirements than existing solutions like Shiki.
Consultant Jeff Greenberg provides a framework for evaluating AI-assisted software tools in professional video workflows to address technical, security, and maintenance risks. The guide outlines five critical questions for users to assess the stability and reliability of AI-generated tools before integration into client projects.
The Phoronix Test Suite has published benchmark performance data for the AMD Ryzen Threadripper 9980X processor using the SVT-AV1 encoder. These metrics provide technical engineers with hardware-specific results to assist in capacity planning and performance optimization for high-density AV1 encoding workflows.
NETINT Technologies explains the operational advantages of using dedicated video processing units (VPUs) over general-purpose CPUs for video encoding pipelines. The technical breakdown details how keeping decode, scaling, and encode stages on a single hardware device reduces data bus contention and increases aggregate throughput for multi-session streaming workflows.
The Unified Patent Court in Düsseldorf has issued an injunction preventing Disney from using specific HEVC bitstreams in eleven European countries due to patent infringement against InterDigital's video coding technologies. The court also established significant procedural guidance regarding the role of non-disclosure agreements in FRAND licensing negotiations.
Interra Systems will showcase new AI-integrated workflows and product updates at IBC 2026, focusing on automating QC, monitoring, and captioning for streaming and broadcast operators. Key updates include expanded support for NDI and Zixi, as well as AI-powered diagnostic and perceptual quality scoring features across the Orion, Baton, and Vega platforms.
Researchers from Meta, Brown University, and MIT have introduced Ms. Forcing, a streaming video generation paradigm that utilizes multi-scale patchification to minimize computational redundancy. The approach achieved a 39.6% speed increase over the Rolling Forcing baseline, reaching 22.84 FPS on a single NVIDIA H200 GPU while improving visual stability for long-horizon generation.
Whisper and the Fédération Française de Tennis (FFT) have scaled their remote production workflow for the 2026 French Open, utilizing a London-based hub to manage 231 staff and provide feeds for 28 broadcast partners. The production update includes new curated behind-the-scenes content and expanded camera and replay infrastructure across the tournament courts.
The global multi-format transcoding device market is forecast to grow from $1.26 billion in 2025 to $2.35 billion by 2034, driven by increasing UHD traffic, 5G rollouts, and the transition to high-efficiency codecs. The market shift reflects a growing demand for cloud-native, AI-enhanced, and real-time transcoding solutions capable of handling AV1, VVC, and immersive media formats.
ToolsOnAir has received official NDI certification for its macOS-based capture and playout product families, including the Just In and Just Play lines. The certification confirms adherence to NDI standards for multichannel recording, live-input switching, and the transport of closed-captioning and SCTE-104 signaling within professional broadcast workflows.
LG Electronics has introduced a 'Creator Original Picture Mode' for its 2026 OLED TV lineup that allows studios to push specific color, motion, and shadow parameters directly to the display via an API in the Amazon Prime Video app. The functionality leverages the advanced image processing of LG’s new Alpha 11 Gen 3 processor to execute these studio-defined settings within a 12-bit pipeline.
Primate Labs has released Geekbench 7, which introduces new specialized workloads for testing AV1 video encoding, Opus audio compression, and machine learning-powered tasks like live captioning and video upscaling. These benchmarks are designed to help media engineers evaluate hardware performance for modern streaming codecs and AI-driven video processing workflows.
Brazil’s broadcast engineering society (SET) and middleware provider Mirakulo have validated the Advanced HDR by Technicolor ecosystem as production-ready for the country's upcoming TV 3.0 standard. The verification, conducted at a Philips certification lab, confirms that current chipsets and hardware are prepared to support widespread HDR deployment over the air.
A developer has released a containerized benchmarking tool for Unraid media servers designed to standardize hardware transcoding performance measurements across Intel, Nvidia, and AMD GPUs. The tool allows users to test server transcoding capacity for various streams and contribute data to a centralized community leaderboard to help operators understand hardware limitations.
AMD researchers introduced a tetrahedral cage-based method to decouple animation costs from triangle density in real-time ray tracing, achieving 60 FPS performance with 585 million animated triangles. The technique utilizes low-resolution proxy meshes to animate complex geometry, reducing memory and computational overhead for virtual production and high-fidelity rendering environments.
Cadence Design Systems has published a technical guide detailing microarchitecture exploration techniques for its Stratus high-level synthesis (HLS) tool. The documentation explains how designers can optimize area, power, and latency tradeoffs by utilizing synthesis attributes, directives, and integration with Cadence Cerebrus machine learning tools without modifying source code.
Developers have identified a reproducible bug in the NVIDIA DeepStream 9.1 SDK where the nvurisrcbin component causes a process abort during RTSP reconnection when NTP synchronization is active. Users have successfully mitigated the issue by replacing the component with uridecodebin, which maintains stability and accurate timing without causing crashes.
A developer has submitted a V4L2 driver for the VEPU510 H.264 hardware encoder found in Rockchip’s RK3576 SoC for inclusion in the Linux kernel main line. While the driver successfully handles intra-frame encoding, it currently encounters hardware watchdog stalls during P-frame processing that require further debugging.
AWS researchers conducted an independent benchmark of NDI SpeedHQ and JPEG XS codecs for live cloud production workflows, testing image fidelity and generational quality loss using PSNR metrics. The study evaluated both codecs on AWS EC2 instances, concluding that while JPEG XS provides higher absolute quality, NDI SpeedHQ performs more efficiently at lower bitrates.
This article provides a technical overview of the differences between Variable Bitrate (VBR) and Constant Bitrate (CBR) encoding for video streaming. It outlines the specific use cases for each, noting that VBR is generally preferred for VOD and archiving while CBR is the industry standard for live-streaming environments.
A developer has identified a performance regression in NVIDIA CloudXR 6.2.x, where the server-side HEVC video encoding pipeline fails to produce frames for the Apple Vision Pro on Linux. The report notes that this issue is specific to the updated runtime and does not occur in version 5.0.1 or with standard FFmpeg NVENC deployments on the same hardware.
NVIDIA documented the implementation of the IProgressMonitor API for its TensorRT engine, enabling developers to observe and cancel long-running build processes in Python or C++. This update aims to assist engineers in optimizing GPU utilization and managing agentic AI workflows by surfacing real-time build telemetry.
Recent benchmarks of AMD EPYC Turin processors running on Linux 7.2 demonstrate performance gains in FFmpeg libx264 encoding and x265 4K video processing compared to the previous kernel version. These results highlight improvements in server efficiency for video-centric compute workloads.
NSFOCUS CERT has disclosed a critical remote code execution vulnerability (CVE-related) in Alibaba's Fastjson 1.2.x library affecting versions between 1.2.68 and 1.2.83. Security specialists are urging developers managing streaming infrastructure to migrate to Fastjson 2.x or enable the SafeMode function to mitigate exploitation risks.
The MPEG-4 Visual Patent Portfolio License has officially terminated following the expiration of its final patent in Brazil on July 19, 2026. This conclusion removes ongoing royalty obligations and participation fees for codec technologies formerly managed by the VIA Licensing Alliance.
Researchers at the University of Notre Dame have introduced EVOLVE, an autoencoder-based framework designed for high-ratio scientific volume compression. Unlike current implicit neural representation methods, EVOLVE supports variable-rate encoding within a single model and offers significantly faster processing speeds for large-scale data sets.
