All articles tagged with #video encoding
FFmpeg 8.0 is set to be released by the end of August, introducing numerous new features including support for new codecs, hardware acceleration, AI-driven audio filtering, and various performance improvements, making it a significant update for multimedia processing.
AMD has added region of interest (ROI) encoding functionality to their open-source Mesa code for Linux, specifically for Video Core Next (VCN) enabled Radeon graphics hardware. This feature allows users to specify areas of the video frame that are higher or lower priority, enabling the video encoder hardware to focus its bit usage for optimal results. The ROI video encode feature is useful for applications like AR/VR and mapping persistent text to specific regions of the frame. The code is available in Mesa Git for the upcoming Mesa 24.0 feature release in Q1, along with improvements to the RadeonSI Gallium3D and RADV Vulkan drivers.
OBS Studio 29.1 now supports AV1 stream encoding when broadcasting to YouTube, enabling higher-quality footage than H.264 at similar bitrates and higher resolutions. The new Enhanced RTMP standard and hardware encoding work with AMD, Intel, and NVIDIA GPUs, and software can be used if the CPU is powerful enough. YouTube still transcodes the output to its VP9 format, but the image quality loss is slight. The change hints at a future where creators can stream video that looks as good to viewers as it does on the host's PC.
Nvidia has quietly updated its support page to reveal that most of its GPUs can now encode up to five video streams simultaneously, up from the previous limit of three. This unlocks capabilities that were always present in the hardware but were software-limited in consumer GPUs. The change applies to most GeForce GPUs going back to the 2014-era Maxwell architecture, and users may need to install the most recent drivers to unlock the additional encoding capabilities. However, video quality settings may still limit the number of video streams that can be encoded simultaneously.