Release of the FFmpeg 9.1 multimedia package

After three months of development, the multimedia package FFmpeg 9.1 has been released, including a set of applications and a collection of libraries for operations on various multimedia formats (recording, conversion, and decoding of audio and video formats). The package is written in C and distributed under LGPL and GPL licenses.

Changes in FFmpeg 9.1 include:

  • A network backend based on the libcurl library has been added.
  • An experimental API has been proposed for creating bitstream filters for direct modification of the packed bitstream without decoding it. The new API allows filters to be combined into non-linear graphs that enable complex routing, merging, and splitting of streams.
  • The command-line utility ffmpeg has been enhanced with the ability to work with dual-track representations of the LCEVC (Low Complexity Enhancement Video Coding) video encoding format, where the base video and quality enhancement metadata are stored as separate tracks.
  • Support has been added for transferring decoded video frames from the NVIDIA NVDEC hardware decoder to GPU memory using CUDA arrays (CUARRAY) instead of the CUdeviceptr buffer to improve compatibility with the Vulkan graphics API. CUARRAY support has been added for H.264, HEVC, AV1, MPEG-1, MPEG-2, MPEG-4, VC-1, WMV3, VP8, and VP9 codecs.
  • Packagers and unpackagers for the astc and KTX 1.0 media containers have been added, used for storing textures. Support for compressing and unpacking textures in the ASTC (Adaptive Scalable Texture Compression) format has been implemented using the libastcenc library.
  • Support for selecting AVFoundation video and audio capture devices by unique identifiers and USB device serial numbers has been implemented. The options '-video_device_id' and '-audio_device_id' have been added for specifying identifiers.
  • A new encoder option (apv_vulkan) for the APV (Advanced Professional Video) format has been added, utilizing the capabilities of the Vulkan graphics API and compute shaders for acceleration.
  • An encoder for the DSD format (dsd_msbf) has been added.
  • The encoder for the AAC (Advanced Audio Coding) audio format has been significantly improved. Support for decoding the xHE-AAC format (Extended High-Efficiency AAC, ISO/IEC 23003-3 standard, used in Netflix streaming) has been added using the libfdk-xhe-aac library.
  • Support for decoding partitioned H.264 video streams, where each frame is divided into several parts, has been added.
  • Decoders have been added for the formats Infinity IMM6, Rhetorex ADPCM, ADPCM IMA Citrix, and Panasonic RW2.
  • Implementations of a decoder and unpacker for the DVD-Audio LPCM (Linear Pulse Code Modulated) media container format have been completed.
  • Packers for media containers in the iTerm2 and JPEG Multi-Picture formats have been added.
  • Unpackers for media containers in MVR and JPEG Multi-Picture formats have been added.
  • Support for the "handle:" protocol has been implemented on the Windows platform, allowing the use of HANDLE descriptors from WinAPI and supporting random access instead of only linear reading as with the "pipe:" protocol.
  • The ONNX Runtime backend has added the ability to execute neural network filters in Windows, utilizing the DirectML library implemented over the Direct3D 12 API in zero-copy mode (without copying video frames between GPU memory and RAM).
  • New filters:
    • vf_fruc_vulkan — increases video frame rates by interpolating intermediate frames using the NVIDIA Optical Flow Accelerator with the Vulkan extension VK_NV_optical_flow.
    • gainmap — computes a brightness enhancement map (HDR Gain Map) based on the SDR and HDR versions of the same image.
    • latticepal — converts full-color RGB/BGR images to a format with an 8-bit palette PAL8.
    • vf_vqe_amf — enhances video quality on AMD GPU systems using the VQE (Video Quality Enhancer) noise suppression mechanism available in the Advanced Media Framework (AMF).
    • Bitstream filters added: trim (removing packets outside a specified range), jpegapp (substituting segments JPEG APP0..APP15), and side_data (removing auxiliary metadata).
    • A layer merging filter (a layer with base video and a layer with metadata for quality enhancement) for video in LCEVC format has been added.

Source: opennet.ru

Buy reliable website hosting with DDoS protection, VPS VDS servers 🔥 Buy reliable website hosting with DDoS protection, VPS VDS servers | ProHoster