Hardware acceleration

Hardware acceleration

Hardware accelerated encoding/decoding can be used in encore if it is supported in the ffmpeg executable used.

Encoding

To use hardware accelerated encoding, prepare an encoding profile with a VideoEncode where the codec property is set to the name of a hardware encoders, such as hevc_nvenc. See example profile below.

Decoding

To use hardware decoding, the appropriate input parameters must be specified in the encore job. See example job below. For more details on input configuration, see the Input Configuration guide.

Hardware filters

Note: Support for hardware filters as described below was added in encore version 0.2.8-4.

Hardware filters can be configured like other filters through the videoFilters property of inputs or the filters property of a video encode. It is up to the user to make sure that hardware filters are fed with hardware frames by either using hardware decoding (see above) or inserting hwupload in the filter chain if necessary.

Some filters are inserted automatically by encore. It is possible to configure a profile so that encore uses equivalent hardware filters instead. Hardware replacements for split, scale, crop, and pad filter can be specified. See example profile below. Specifying hardware equivalents for scale, crop, and pad filters are useful, as encore will then use the hardware filters to scale/crop/pad according to dimensions and aspect ratios configured in the job and/or encoding profiles.

Hardware split filter

Specifying a hardware equivalent for the split filter is not useful in most cases, since the standard split filter has no problem handling hardware frames. However, some hardware supports custom split filter with extended functionality.

The splitFilter property will be treated differently depending on if the values contains a '=' or not. If no '=' is included, the value is treated as the name of the filter to use and something like 'SPLITFILTERVALUE=N[out1][out2]...' will be added to the filtergraph, where N is the number of relevant outputs in the profile. If an '=' is included, the value is assumed to already include the size parameters and something like 'SPLITFILTERVALUE[out1][out2]...' will be added to the filtergraph. Care must be taken in this case to ensure that the size parameters match the number of relevant outputs in the profile. This latter form of specifying the split filter can be useful for certain custom split filters that allow extra parameters, ie ni_quadra_split filter for netint quadra cards which allows access to scaled output from the decoder.

Examples

Cuda/nvenc encoding profile

name: nvenc-test
description: "NVENC HEVC profile

Note:
This profile requires encore version 0.2.8-4 or later.
This profile requires using cuda hardware decoding by specifying cuda
hardware decoding on the input
by setting parameters as below
    \"params\": {
        \"hwaccel\": \"cuda\",
        \"hwaccel_output_format\": \"cuda\"
      }
"
scaling: bicubic
filterSettings:
  scaleFilter: scale_cuda
  scaleFilterParams:
    format: p010le
  padFilter: pad_cuda
encodes:
  - type: VideoEncode
    codec: hevc_nvenc
    format: mp4
    suffix: _h265_2600
    twoPass: false
    height: 1080
    params:
      b:v: 2600k
      maxrate: 3900k
      bufsize: 5200k
      r: 25
      profile:v: main10
      tag:v: hvc1
      force_key_frames: expr:not(mod(n,96))
    audioEncode:
      type: AudioEncode
      codec: aac
      bitrate: 192k
      suffix: STEREO

Cuda/nvenc encore job

Example job below using cuda hardware acceleration for decoding, and hardware decoding (hevc_nvenc) through the encoding profile above.

{
  "profile": "nvenc-test",
  "outputFolder": "/output/folder/example/",
  "baseName": "example-hw-acceleration",
  "inputs": [
    {
      "uri": "/videos/example.mp4",
      "type": "AudioVideo",
      "params": {
        "hwaccel": "cuda",
        "hwaccel_output_format": "cuda"
      }
    }
  ]
}