How to Use scale_cuda Filter in FFmpeg

This guide explains how to use the hardware-accelerated scale_cuda filter in FFmpeg to resize videos using NVIDIA GPUs. You will learn the basic syntax, system requirements, and practical command-line examples for integrating CUDA-based scaling into your video processing workflows.

Prerequisites

To use the scale_cuda filter, your system must meet the following requirements: * An NVIDIA graphics card that supports CUDA. * NVIDIA proprietary drivers installed. * A build of FFmpeg compiled with CUDA support (specifically enabled with --enable-cuda-nvcc and --enable-libnpp).

The Basic Command Structure

Because scale_cuda runs on the GPU, the video frames must reside in GPU memory (VRAM) before the filter can process them. The most efficient way to use this filter is within a fully hardware-accelerated pipeline, where decoding, scaling, and encoding all happen on the GPU.

Here is the standard command for a full hardware pipeline:

ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input.mp4 -vf scale_cuda=1920:1080 -c:v h264_nvenc output.mp4

Command Breakdown:

• -hwaccel cuda: Enables CUDA hardware acceleration for decoding.
• -hwaccel_output_format cuda: Keeps the decoded video frames in GPU memory, preventing costly transfers between system RAM and GPU VRAM.
• -vf scale_cuda=1920:1080: Invokes the CUDA scaling filter to resize the video to a width of 1920 pixels and a height of 1080 pixels.
• -c:v h264_nvenc: Uses the NVIDIA hardware-accelerated H.264 encoder to compress the scaled video.

Advanced Usage and Syntax

Maintaining Aspect Ratio

You can scale a video while preserving its original aspect ratio by setting one of the dimensions to -1 or -2 (which ensures the dimension is divisible by 2, required by many encoders):

ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input.mp4 -vf scale_cuda=1280:-2 -c:v h264_nvenc output.mp4

Scaling Software-Decoded Video (Hybrid Pipeline)

If your input video format cannot be decoded by your GPU, you must decode it using the CPU, upload the frames to the GPU using the hwupload_cuda filter, apply the scaling, and then encode:

ffmpeg -i input.mp4 -vf "hwupload_cuda,scale_cuda=1280:720" -c:v h264_nvenc output.mp4

Specifying Interpolation Algorithm

The scale_cuda filter supports different scaling algorithms (such as bilinear, bicubic, or lanczos) to balance speed and visual quality. You can specify the algorithm using the interp option:

ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input.mp4 -vf scale_cuda=1920:1080:interp=lanczos -c:v h264_nvenc output.mp4