NVIDIA Video Codec SDK

https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/
Admin
NVIDIA Video Codec SDK provides APIs and tools for hardware-accelerated video encoding and decoding...
Tokens:93,667
Snippets:443
Trust Score:10
Update:1 day ago
Show doc for...
Tokens:
Raw
### Make and install for Linux

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

After configuring with CMake, run 'make' and then 'make install' to build and install the sample application binaries. The binaries will be available in Samples/build.

```make
make
```

```make
make install
```

--------------------------------

### Install ffnvcodec

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/ffmpeg-with-nvidia-gpu/index.html

Installs the nv-codec-headers after cloning.

```bash
cd nv-codec-headers && sudo make install && cd –
```

--------------------------------

### Build samples in release mode with CMake and Make

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Run these commands from the 'build' subfolder inside Video_Codec_SDK_x.y.z/Samples to compile and install the sample applications. The binaries will be placed in 'Samples/build'. Applications using OpenGL and Vulkan will not be built.

```bash
cmake -DCMAKE_BUILD_TYPE=Release ..
```

```bash
make
```

```bash
make install
```

--------------------------------

### Install FFmpeg libraries

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/ffmpeg-with-nvidia-gpu/index.html

Installs the compiled FFmpeg libraries.

```bash
sudo make install
```

--------------------------------

### Build the solution with CMake

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Use this command to build the generated Visual Studio solution and install the sample applications. The binaries will be available in Samples/build.

```cmake
cmake --build . --target install --config Release
```

--------------------------------

### AppTransOneToN sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppTransOneToN sample application, which performs 1:N transcoding of a single input stream. If no output resolutions are specified, it generates two streams of 1280x720 and 800x480.

```bash
AppTransOneToN.exe -i input.h264 -o out -r 1280x720 800x480
```

--------------------------------

### Install necessary packages for FFmpeg compilation

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/ffmpeg-with-nvidia-gpu/index.html

Installs build dependencies required for compiling FFmpeg on Linux.

```bash
sudo apt-get install build-essential yasm cmake libtool libc6 libc6-dev unzip wget libnuma1 libnuma-dev
```

--------------------------------

### NvEncGetEncodeCaps - Query Temporal SVC Support

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/nvenc-video-encoder-api-prog-guide/index.html

Queries the encoder capabilities to check if temporal SVC is supported and to get the maximum number of temporal layers.

```APIDOC
## NvEncGetEncodeCaps

### Description
Queries the encoder capabilities to determine if temporal SVC is supported and to retrieve the maximum number of temporal layers.

### Method
NvEncGetEncodeCaps

### Parameters
- **NV_ENC_CAPS** (enum) - Required - The capability to query. Use `NV_ENC_CAPS_SUPPORT_TEMPORAL_SVC` to check support and `NV_ENC_CAPS_NUM_MAX_TEMPORAL_LAYERS` to get the maximum number of temporal layers.

### Response
Returns a capability value indicating support or the maximum number of temporal layers.

### Example
```cpp
NV_ENC_CAPS_PARAM capsParam = {};
capsParam.capsToQuery = NV_ENC_CAPS_SUPPORT_TEMPORAL_SVC;
int support = nvEncGetEncodeCaps(encoder, &capsParam);
```
```

--------------------------------

### AppEncD3D12 sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppEncD3D12 sample application, which encodes ID3D12Resource. Supports RGB input.

```bash
AppEncD3D12.exe -i input.bgra -s 1920x1080 -o out.h264
```

--------------------------------

### Setup decoder for scaling with cuvidCreateDecoder

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/nvdec-video-decoder-api-prog-guide/index.html

Shows how to configure CUVIDDECODECREATEINFO for scaling from 1280x960 to 1920x1080 before calling cuvidCreateDecoder. The snippet sets ulTargetWidth and ulTargetHeight to the desired output dimensions and then creates the decoder.

```cpp
// Scaling. Source size is 1280x960. Scale to 1920x1080.
CUresult rResult;
unsigned int uScaleW, uScaleH;
uScaleW = 1920;
uScaleH = 1080;
...
CUVIDDECODECREATEINFO stDecodeCreateInfo;
memset(&stDecodeCreateInfo, 0, sizeof(CUVIDDECODECREATEINFO));
... // Setup the remaining structure members
stDecodeCreateInfo.ulTargetWidth  = uScaleWidth;
stDecodeCreateInfo.ulTargetHeight = uScaleHeight;
rResult = cuvidCreateDecoder(&hDecoder, &stDecodeCreateInfo);
...
```

--------------------------------

### AppEncMultiInstance.exe command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppEncMultiInstance sample application, which splits input video into independent GOPs for parallel encoding. Use -nf to set the number of portions and -thread to set the number of encoding session threads. Note that on GeForce GPUs, simultaneous encode sessions are restricted to 5.

```bash
AppEncMultiInstance.exe -i input.yuv -s 1920x1080 -if iyuv -nf 4 -thread 2 -o out.h264
```

--------------------------------

### AppEncD3D11 sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppEncD3D11 sample application, which encodes frames from ID3D11Texture2D textures. Supports RGB input by default and NV12 via the -nv12 parameter.

```bash
AppEncD3D11.exe -i input.bgra -s 1920x1080 -nv12 -o out.h264
```

--------------------------------

### Run AppDecMultiInput sample application

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Decodes multiple raw video files and post-processes them with CUDA kernels on different CUDA streams. Applies a Ripple effect as part of post-processing.

```plaintext
AppDecMultiInput.exe -i input1.yuv -s 1920x1080 -i input2.yuv -s 1280x720
```

--------------------------------

### AppEncD3D9 sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppEncD3D9 sample application, which encodes frames from IDirect3DSurface9 surfaces. Supports RGB input by default and NV12 via the -nv12 parameter.

```bash
AppEncD3D9.exe -i input.bgra -s 1920x1080 -nv12 -o out.h264
```

--------------------------------

### Run AppDecMultiFiles sample application

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Decodes multiple files sequentially with or without using the decoder Reconfigure API. Displays time taken for decoder creation and destruction. Files are specified using -filelist.

```plaintext
AppDecMultiFiles.exe -filelist files.txt
```

--------------------------------

### AppTransPerf sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppTransPerf sample application, which measures transcoding performance in FPS. It takes a single input stream and spawns N pairs of threads for decoding and encoding.

```bash
AppTransPerf.exe -i input.h264 -thread 2
```

--------------------------------

### AppDecD3D sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command-line usage for the AppDecD3D sample application, which decodes a media file and displays frames in a window using CUDA interop with D3D9 or D3D11.

```text
AppDecD3D.exe -i input.h264 -d3d 11
```

--------------------------------

### AppEncDec sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppEncDec sample application, which demonstrates encoding and streaming with one thread while another thread decodes. Supports multiple input formats including IYUV, NV12, NV16, P010, P210, BGRA, and BGRA64.

```bash
AppEncDec.exe -i input.h264 -o out.h264
```

--------------------------------

### Run AppDecPerf sample application

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Measures decoding performance in FPS. Creates multiple host threads, each running a separate decoding session. Default is 2 threads. Supports measuring decode performance only or including transfer to host memory.

```plaintext
AppDecPerf.exe -i input.h264 -thread 2
```

--------------------------------

### AppEncGL sample command line

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Command line for the AppEncGL sample application, which encodes frames stored in OpenGL textures. Requires the X server to be running and the DISPLAY environment variable set. Supports IYUV and NV12 input formats.

```bash
AppEncGL.exe -i input.yuv -s 1920x1080 -if iyuv -o out.h264
```

--------------------------------

### Run AppDecMem sample application

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/read-me/index.html

Demuxes and decodes media content from a memory buffer. Allocates AVIOContext explicitly and defines a method to read data packets from the input file.

```plaintext
AppDecMem.exe -i input.h264
```

--------------------------------

### NVENC External ME Hints Configuration

Source: https://docs.nvidia.com/video-technologies/video-codec-sdk/13.0/nvenc-video-encoder-api-prog-guide/index.html

Steps to enable and configure external ME hints for H.264, HEVC, and AV1 encoders using NVENC initialization and per-frame parameters.

```APIDOC
## NVENC External ME Hints Configuration

### Description
Configure NVENC to accept external motion vector hints for guiding motion search. This applies to H.264, HEVC, and AV1 encoders.

### Initialization Parameters
- **NV_ENC_INITIALIZE_PARAMS::enableExternalMEHints** (boolean) - Required - Set to enable external ME hints.
- **NV_ENC_INITIALIZE_PARAMS::maxMEHintCountsPerBlock** (array) - Required - Maximum number of hint candidates per block per direction for each partition type. Index 0 for L0 predictors, index 1 for L1 predictors.

### Per-Frame Parameters
- **NV_ENC_PIC_PARAMS::meHintCountsPerBlock** (array) - Required - Number of hint candidates per block per direction for different partition types.
- **NV_ENC_PIC_PARAMS::meExternalHints** (pointer) - Required (H.264/HEVC) - Pointer to buffer storing hints.
- **NV_ENC_PIC_PARAMS::meExternalSbHints** (pointer) - Required (AV1) - Pointer to buffer storing SB hints.
- **NV_ENC_PIC_PARAMS::meSbHintsCount** (integer) - Required (AV1) - Total number of external ME SB hint candidates for the frame.

### Hint Structures
- **NVENC_EXTERNAL_ME_HINT** - Used for H.264 and HEVC.
- **NVENC_EXTERNAL_ME_SB_HINT** - Used for AV1.

### Hint Ordering
- For H.264/HEVC: For each MB/CTU, hints must be provided in partition order: 16x16, 16x8, 8x16, 8x8. For each partition type, L0 hints followed by L1 hints (if available).
- For AV1: For each SB, L0 hints followed by L1 hints (if available).

### Supported Partition Sizes
- H.264: 16x16, 16x8, 8x16, 8x8
- HEVC: 16x16 and 8x8 only
- AV1: SB (superblock) based

### Hint Ranges (full pixel units)
- H.264/HEVC: Horizontal [-2048, 2047], Vertical [-512, 511]
- AV1: Horizontal [-1023, 1023], Vertical [-511, 511]

### Usage Steps
1. Set `enableExternalMEHints` and `maxMEHintCountsPerBlock` during initialization.
2. For each frame, set `meHintCountsPerBlock` and provide hint buffer pointer (`meExternalHints` for H.264/HEVC, `meExternalSbHints` for AV1). For AV1, also set `meSbHintsCount`.
3. Provide hints in the specified order and range.
```
Tokens:
Raw