Skip to content

SVT HEVC encoder. Scalable Video Technology (SVT) is a software-based video coding technology that is highly optimized for Intel® Xeon® processors. Using the open source SVT-HEVC encoder, it is possible to spread video encoding processing across multiple Intel® Xeon® processors to achieve a real advantage of processing efficiency.

License

Notifications You must be signed in to change notification settings

jonrecker/SVT-HEVC

 
 

Repository files navigation

Scalable Video Technology for HEVC Encoder (SVT-HEVC Encoder)

AppVeyor Build Status Travis Build Status Coverage Status

The Scalable Video Technology for HEVC Encoder (SVT-HEVC Encoder) is an HEVC-compliant encoder library core that achieves excellent density-quality tradeoffs, and is highly optimized for Intel® Xeon™ Scalable Processor and Xeon™ D processors.

The whitepaper for SVT-HEVC can be found here: https://01.org/svt

This encoder has been optimized to achieve excellent performance levels using 12 density-quality presets (please refer to the user guide for more details).

License

Scalable Video Technology is licensed under the OSI-approved BSD+Patent license. See LICENSE for details.

Documentation

More details about the encoder usage can be found under:

System Requirements

Operating System

SVT-HEVC may run on any Windows* or Linux* 64 bit operating systems. The list below represents the operating systems that the encoder application and library were tested and validated on:

  • Windows* Operating Systems (64-bit):

    • Windows Server 2016
  • Linux* Operating Systems (64-bit):

    • Ubuntu 16.04 Server LTS

    • Ubuntu 18.04 Server LTS

    • CentOS 7.4/7.5/7.6

Hardware

The SVT-HEVC Encoder library supports x86 architecture

  • CPU Requirements

In order to achieve the performance targeted by the encoder, the specific CPU model listed above would need to be used when running the encoder. Otherwise, the encoder runs on any 5th Generation Intel Core™ Processors (formerly Broadwell) CPUs (Xeon E5-v4) or newer.

  • RAM Requirements

In order to run the highest resolution supported by the encoder, at least 64GB of RAM is required to run a single 8kp50/10-bit encode. The encoder application will display an error if the system does not have enough RAM to support such. The following table shows the minimum amount of RAM required for some standard resolutions of 10bit video per stream:

Resolution Minimum Footprint (GB)
8k 64
4k 16
1080p 6
720p/1080i 4
480p 3

Build and Install

Windows* Operating Systems (64-bit)

  • Build Requirements

    • Visual Studio* 2017 (download here) or 2019 (download here)
    • CMake 3.14 or later (download here)
    • YASM Assembler version 1.2.0 or later
      • Download the yasm exe from the following link
      • Rename yasm-1.3.0-win64.exe to yasm.exe
      • Copy yasm.exe into a location that is in the PATH environment variable
  • Build Instructions

    • Build the project by following the steps below in a windows command prompt:
      • In the main repository directory, cd to <repo dir>\Build\windows
      • Run build.bat [2019|2017|2015] [This will automatically generate and build the project]
    • To Build the project using a generator other than Visual Studio
      • run build.bat [ninja|msys|mingw|unix] instead of the second command
      • Note: These are not officially supported and thus are not displayed in the help message.
  • Binaries Location

    • Binaries can be found under <repo dir>\Bin\Release or <repo dir>\Bin\Debug, depending on whether Debug or Release was selected
  • Installation
    For the binaries to operate properly, the following conditions have to be met:

    • On any of the Windows* Operating Systems listed in the OS requirements section, install Visual Studio* 2017 or 2019
    • Once the build is complete, copy the binaries to a location making sure that both the application SvtHevcEncApp.exe and library SvtHevcEnc.dll are in the same folder.
    • Open the command prompt at the chosen location and run the application to encode. SvtHevcEncApp.exe -i [in.yuv] -w [width] -h [height] -b [out.265]
    • The application also supports reading from pipe. E.g. ffmpeg -i [input.mp4] -nostdin -f rawvideo -pix_fmt yuv420p - | SvtHevcEncApp.exe -i stdin -n [number_of_frames_to_encode] -w [width] -h [height]

Linux* Operating Systems (64-bit)

  • Build Requirements

    • GCC 5.4.0 or later
    • CMake 3.5.1 or later
    • YASM Assembler version 1.2.0 or later
  • Build Instructions

    • In the main repository, run either the provided build script

      cd Build/linux
      ./build.sh [release|debug] [static|shared] [install]
      # Requires sudo permission for installing
      # Run './build.sh -h' to see the full help
    • or run the commands directly

      mkdir build && cd build && cmake .. -DCMAKE_BUILD_TYPE=<Release|Debug> && make -j $(nproc) && sudo make install
  • Sample Binaries location

    • Binaries can be found under $REPO_DIR/Bin/Release
  • Installation
    For the binaries to operate properly, the following conditions have to be met:

    • On any of the Linux* Operating Systems listed above, copy the binaries under a location of your choice.
    • Change the permissions on the sample application “SvtHevcEncApp” executable by running the command: chmod +x SvtHevcEncApp
    • cd into your chosen location
    • Run the sample application to encode. ./SvtHevcEncApp -i [in.yuv] -w [width] -h [height] -b [out.265]
    • Sample application supports reading from pipe. E.g. ffmpeg -i [input.mp4] -nostdin -f rawvideo -pix_fmt yuv420p - | ./SvtHevcEncApp -i stdin -n [number_of_frames_to_encode] -w [width] -h [height]

How to evaluate by ready-to-run executables with docker

Refer to the guide here.

Demo features and limitations

  • VBR BRC mode:
    The VBR functionality implemented in SVT-HEVC Encoder is a demo feature to allow for an easier integration of product level BRC. The algorithm implemented would allow the encoder to generate an output bit stream matching, with a best effort, the target bitrate. The algorithm does not guarantee a certain maximum bitrate or maximum buffer size [does not follow HRD compliance]. When set to encode in VBR mode, the encoder does not produce a bit-exact output from one run to another.

  • Speed Control output:
    The speed control functionality implemented for SVT-HEVC Encoder is a demo feature showcasing the capability of the library to adapt to the resources available on the fly in order to generate the best possible video quality while maintaining a real-time encoding speed. When set to use the Speed Control mode, the encoder does not produce a bit-exact output from one run to another.

  • Multi-instance support:
    The multi-instance functionality is a demo feature implemented in the SVT-HEVC Encoder sample application as an example of one sample application using multiple encoding libraries. Encoding using the multi-instance support is limited to only 6 simultaneous streams. For example two channels encoding on Windows: SvtHevcEncApp.exe -nch 2 -c firstchannel.cfg secondchannel.cfg

How to Contribute

We welcome community contributions to the SVT-HEVC Encoder. Thank you for your time! By contributing to the project, you agree to the license and copyright terms therein and to the release of your contribution under these terms.

Contribution process

  • Follow the coding_guidelines
  • Validate that your changes do not break a build
  • Perform smoke tests and ensure they pass
  • Submit a pull request for review to the maintainer

How to Report Bugs and Provide Feedback

Use the Issues tab on Github

IRC

#svt on Freenode. Join via Freenode Webchat or use your favorite IRC client.

Notices and Disclaimers

The notices and disclaimers can be found here

About

SVT HEVC encoder. Scalable Video Technology (SVT) is a software-based video coding technology that is highly optimized for Intel® Xeon® processors. Using the open source SVT-HEVC encoder, it is possible to spread video encoding processing across multiple Intel® Xeon® processors to achieve a real advantage of processing efficiency.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • C 97.3%
  • Assembly 0.8%
  • C++ 0.8%
  • Python 0.7%
  • CMake 0.3%
  • Shell 0.1%