Check the projects GitHub Page
An easy to use Application to stabilize shaky video footage preferably using Gpu acceleration via Cuda or OpenCL. Cuvista will run without a dedicated Gpu and produce the same output, but will take significantly longer to do so.
I put in great effort to ensure that on all platforms the stabilization results are absolutely equal. By that I mean floating point equal, to the bit.
There is no AI involved, just math, algorithms and brain power.
Have a look at a sample video comparing an original recording side by side to the stabilized version:
For Cuda acceleration a device with Compute Version 7 or later is required. For OpenCL the device must support at least version 2.
You can get the application ready to use in three ways
- From the Microsoft App Store via CUVISTA
- From GitHub on the Releases page you can download an msi file to install the latest version
- Also on the Releases page you will find a zip file. This version does not need any installation. Download, unzip and run either
cuvista.exeon the command line orcuvistaGui.exefor a graphical user interface. This version might require you to install additional components such as the Microsoft Visual C/C++ Runtime Libraries which you find at https://aka.ms/vs/17/release/vc_redist.x64.exe
Have a look at available options on the command line via cuvista -h or cuvista -help, in the GUI version a subset of most important options is available.
To start, get the repository and submodules to your computer
git clone --recursive https://github.com/RainerMtb/cuvista.git
This repo comes with a cmake script. As the most bare version only the command line executable is built without Cuda support.
The build script will look for Cuda and include Cuda when found. Tell cmake where to find Cuda by setting CMAKE_PREFIX_PATH=path/to/cuda or via the system path. Explicitly disable Cuda in the build via option -D BUILD_CUDA=0. Explicitly enable Cuda via -D BUILD_CUDA=1, this way skipping the search process
The build script will look for Qt6 and include the Gui executable when found. Tell cmake where to find Qt6 by setting CMAKE_PREFIX_PATH=path/to/qt or via the system path. Explicitly disable building the gui via option -D BUILD_GUI=0
Tested on Windows 11 and MSVC
Get ffmeg with shared libraries https://www.ffmpeg.org/download.html#build-windows
Optionally get Cuda https://developer.nvidia.com/cuda-downloads
Optionally get Qt6 https://www.qt.io/download-qt-installer, install at least the essentials and Qt Multimedia
In a command line window, starting from the project main directory, where this readme and the top level CMakeLists.txt is located, exectute the following commands.
It is recommended to build in a subdirectory:
mkdir build
cd build
prepare the locations of Cuda, Qt and FFMPEG and provide them to cmake. Also see notes above. Better use forward slashes / to separate folders. Adapt locations to your system:
cmake .. -D CMAKE_PREFIX_PATH=C:/CUDA/version;C:/Qt/6.11.1/msvc2022_64 -D FFMPEG_PATH=C:/ffmpeg --fresh
cmake --build . --config Release
Upon successfull completion you will get the files
cuvistaCli/Release/cuvista.exe
cuvistaGui/Release/cuvistaGui.exe
Optionally continue with:
cmake --install . --config Release
This will put together the dependent libraries and the executables in the subfolder install
When done, run cuvista -info to see available devices and run a test
Tested on Ubuntu 26.04
Optionally get Cuda https://developer.nvidia.com/cuda-downloads and follow instructions there
When using Cuda you will also need Nvidia Video encoder - I wonder why that library carries the name ffmpeg...
sudo apt install -y libffmpeg-nvenc-dev
Optionally get Qt6 to build a Gui, see https://www.qt.io/download-qt-installer. To get the necessary components on the command line
wget https://download.qt.io/official_releases/online_installers/qt-online-installer-linux-x64-online.run
chmod +x qt-online-installer-linux-x64-online.run
./qt-online-installer-linux-x64-online.run install qt.qt6.6111.linux_gcc_64 qt.qt6.6111.addons.qtmultimedia
This will then require
sudo apt install -y libxkbcommon-dev libglu1-mesa-dev
Cmake, the build tool:
sudo apt install -y cmake
FFmpeg shared libraries:
sudo apt install -y libavcodec-dev libavdevice-dev libavfilter-dev libavformat-dev libavutil-dev libswresample-dev libswscale-dev
PkgConfig is used to check ffmpeg components:
sudo apt install -y pkg-config
In a terminal session, starting from the project main directory, where this readme and the top level CMakeLists.txt is located
It is recommended to build in a subdirectory:
mkdir build
cd build
Adapt the location of Cuda and Qt to your system and execute the build process. Also see notes above.
export CMAKE_PREFIX_PATH=~/Qt/6.11.1/gcc_64:/usr/local/cuda
cmake .. --fresh
cmake --build . --parallel
Upon successfull completion you will find the executables
cuvistaCli/cuvista
cuvistaGui/cuvistaGui
cuvistaFFmpeg/libcuvistaFFmpeg.so
The executables need to dynamically load the library, in order make the operating system find the file you may use
export LD_LIBRARY_PATH=/path/to/cuvistaFFmpeg
Finally run cuvistaCli/cuvista -info to check your setup
- Improve performance and robustness of stabilization
- Make a plugin for video editing software
- Implement Vulkan Compute
- Windows 11
- Ubuntu 26.04
- Visual Studio 2026
- Nvidia Cuda 13.3.1
- Nvidia Video Codec SDK 13.0.37
- FFmpeg 8.1
- Qt 6.11.2
- Cmake later than 3.28