SDK Package

• Merged the Lite SDK package into the Standard SDK package to provide a unified distribution.

Platform

• Add support for deployment in Docker (Linux guest system).

Samples

• Replace the CMake option -DSPINSDK_LINUX_LIB with -DSPINSDK_LIB_SUB_DIR for specifying library variants on Linux.

Stream CLI apps

• Rework framework and implementation to allow for low latency
  • Global –lowlat option
• Rework pipeline setup to allow for easier multi-input multi-output elements
• Fix NVDEC dts generation
• Add LibAV hardware decoder support (–ffhwdec, linux only)
• Update to NVCodec SDK 13.0.19
• Moved the NVCodec-supported versions of streamenc and streamplay to the standard SpinSDK package as streamenc_nv and streamplay_nv.
• streamenc
  • Add support for rendering preview
  • Add spincompare support
• streamplay
  • Add support for stream sink (–sout) to remux and store incoming streams
  • Fix logo overlay toggling
  • Linked zoom with multiple renderer windows

License

• Fix crash when SpinLib_InitLicense is called with a null pointer

SpinCommon

• Add support for the BC5 formats (SE_PF_BC5_420, SE_PF_BC5_422)
• Add SE_STEREO3D_Left and SE_STEREO3D_Right to identify views
• Add Spin_ExtVideoFormat to describe multilayer video
• Remove deprecated SD_E_INVALIDPICTURE, use SD_E_INVALID_PICTURE instead
• Add SD_E_INVALID_ARGS error code
• Extend SE_SEI enum types

SpinEnc

• Add support for MV-HEVC, compatible with Apple Vision Pro
• Add support for low latency encoding (1 frame latency).
  • Lookahead and encoding stage now share the combined number of frames-in-flight
  • Improve rate control in low latency conditions (small lookahead and small buffer scenarios)
• Add support for GDR (experimental)
• Add support for multilayer VVC (multiview and scalable coding, experimental)
• Add ITUT-T35 user registered SEI support
  • Add SpinSEI_UserDataRegisteredT35 struct
• Add support for per frame HDR SEI metadata
• Add support for inserting dynamic random access points (CRA/IDR)
• Add support for output of the decoded picture together with the compressed picture
• Add codec Level and Tier override option
• Add MCTF enable/disable control
• Add more fine-grained perceptual mode controls
  • SpinEnc_Param::Perceptual::fPerceptStrength to attenuate the overall factor strength factor
  • SpinEnc_Param::Perceptual::fPerceptTempStr to attenuate the strength of the temporal component
  • SpinEnc_Param::Perceptual::fPerceptPenStr to attenuate the strength of penalties
• New screen content detector and handling
  • Significantly better classification
  • Disables MCTF for screen content
  • Disable some perceptual optimization for screen content
  • Use default flat quantization matrix with screen content
• Fix intra mode RDO on 64×64 block level
• Add SD_E_BUFFER_LAYER_MISMATCH error code
• API changes
  • Restructure SpinEnc_Descript with SpinEnc_Descript::LayerDescript to allow for multilayer description
  • Rename SpinEncLib_EncodeFrame to SpinEncLib_EncodePicture
  • Add SpinEncLib_MapPicture2 and SpinEncLib_EncodePicture2 to allow multilayer encoding
    • Old API still available but without new capabilities (SpinEncLib_EncodeFrame, SpinEncLib_MapPicture)
  • Extend SpinEncLib_CreateSEINal to allow for layer and temporal layer configuration
  • Change Spin_VideoFormat to Spin_ExtVideoFormat for sEncInFmt to allow for multilayer encoding input config
  • Rearrange some options to allow specifiying them per layer SpinEnc_Param::LayerOptions

SpinDec

• Reduce latency SpinDecLib_DecodeFrame interface
  • Assume user supplied entire compressed frame data and start decoding immediately
• Fix inferring number of sublayer MV-HEVC
• Add SD_E_INVALID_SLICE error code
• Add layerId and maxNumReorder to SpinDec_Descriptor

SpinFilter

• Introduce the merge filter (SE_FILT_Merge) for combining multiple inputs to single output image.
• Replace and extend split filter (SE_FILT_Split) with the new expand filter (SE_FILT_Expand).
  • Implement clone (SE_OUT_Clone) output mode
  • Add stereo split (SE_OUT_StereoSplit) outout mode
• Introduce a new filter (SE_FILT_BlurSharp) for blurring and sharpening
• Fix cropping filter when intermediate results is used for output
• API changes
  • Replace Spin_Picture sFiltInFmt to Spin_ExtVideoFormat sInFmt to allow for multiple inputs

SpinRender

• Add support for collecting E2E latency stats (time until final present call)
• Add cx and cy option to specify center of overlay
• Add tracking of end-to-end latency in SpinRender_StatConf
• Disable default high accuracy timing option
• Fix wrong aspect ratio black bars appearing when zooming out
• Fix zoom low latency renderer
• Use single device zooming in non-tiled modes
• Adjust overlay sizing for videos with height > width

SpinRawIO

• Add support for Windows Media Foundation capture sources (SE_IODEV_MFCAPTURE)

SpinStream

• Add low latency streaming support
  • Incoming packets are demuxed as soon as possible
  • Outgoing frames are muxed and sent as soon as possible
  • Currently supported for MPEG-TS using RTP sender and receiver
  • Update MPEG-TS (de)muxing to use PS packet size to reduce latency
• Add tracking of end-to-end latency in SpinStream_StatConf
• Add option to output SDP file when muxing RTP streams
• Add bRealTime param to simulate real-time sources with files
• Add support for the Caton streaming protocol
• Add SD_E_SRT_BIND, SD_E_CATON_MISSING_ID, SD_E_CATON_MISSING_KEY error codes
• API changes
  • Add SpinStreamLib_GetPacket2 to allow for non-blocking and timer based packet retrieval
  • Rename “muxrate” to “maxrate” in the url options

SpinCompare

• Add VMAF support (linux only, NEG model, 1080P/4K autoselect)