Video stabilization using gyroscope data

Homepage • Download • Documentation • Discord • Report bug • Request feature

About the project

Gyroflow is an application that can stabilize your video by using motion data from a gyroscope and optionally an accelerometer. Modern cameras record that data internally (GoPro, Sony, Insta360 etc), and this application stabilizes the captured footage precisely by using them. It can also use gyro data from an external source (eg. from Betaflight blackbox).

Trailer / results video

Features

• Real time preview, params adjustments and all calculations
• GPU processing and rendering
• Fully multi-threaded
• Rolling shutter correction
• Supports already stabilized GoPro videos (captured with Hypersmooth enabled) (Hero 8 and up)
• Supports and renders 10-bit videos (up to 16-bit 4:4:4:4 for regular codecs and 32-bit float for OpenEXR - works directly on YUV data to keep maximum quality)
• Customizable lens correction strength
• Render queue
• Visual chart with gyro data (can display gyro, accel, magnetometer and quaternions)
• Visual display of smoothed quaternions
• Real time offset adjustments
• Two optical flow methods
• Two offsets calculation methods
• Modern responsive user interface with Dark and Light theme
• Adaptive zoom (dynamic cropping)
• Support for image sequences (PNG and OpenEXR)
• Based on telemetry-parser - supports all gyro sources out of the box
• Gyro low pass filter, arbitrary rotation (pitch, roll, yaw angles) and orientation
• Multiple gyro integration methods for orientation determination
• Multiple video orientation smoothing algorithms, including horizon levelling and per-axis smoothness adjustment.
• Cross-platform - currently works on Windows/Linux/Mac, with Android and iOS apps coming
• Multiple UI languages
• Supports variable and high frame rate videos, all calculations are done on timestamps
• x264, x265, ProRes, DNxHD, PNG and OpenEXR outputs, with x264 and x265 fully GPU accelerated (ProRes also accelerated on Apple M1 Pro/Max/Ultra)
• Automatic lens calibration process
• Fully zero-copy GPU preview rendering (working but experimental)
• Core engine is a separate library without external dependencies (no Qt, no ffmpeg, no OpenCV), and can be used to create OpenFX and Adobe plugins (on the TODO list)
• Automatic updates of lens profile database
• Built-in lens profiles for GoPro HERO 6, 8, 9 and 10 in all shooting modes

Supported gyro sources

• GoPro (HERO 5 and later)
• Sony (a1, a7c, a7r IV, a7 IV, a7s III, a9 II, FX3, FX6, FX9, RX0 II, RX100 VII, ZV1, ZV-E10)
• Insta360 (OneR, OneRS, SMO 4k, Go, GO2)
• Betaflight blackbox (CSV and binary)
• Runcam CSV (Runcam 5 Orange, iFlight GOCam GR, Runcam Thumb, Mobius Maxi 4K)
• Hawkeye Firefly X Lite CSV
• WitMotion (WT901SDCL binary and *.txt)
• iOS apps: Sensor Logger, G-Field Recorder, Gyro
• Android apps: Sensor Logger, Sensor Record
• Gyroflow .gcsv log
• ArduPilot VideoStabilization logging (*.log)

Info for cameras not on the list

• For cameras which do have built-in gyro, please contact us and we will implement support for that camera. Refer to the documentation for information about the gyro logging process.
• For cameras which don't have built-in gyro, please consider using Betaflight FC or check out our flowshutter project.

Minimum system requirements:

• Windows 10 64-bit (1809 or later) Install VC redist if it doesn't run
  • If you have Windows "N" install, go to Settings -> Apps -> Optional features -> Add a feature -> enable Media Feature Pack
• macOS 10.14 or later (both Intel and Apple Silicon are supported natively)
• Linux:
  • .tar.gz package (recommended): Debian 10+, Ubuntu 18.10+, CentOS 8.2+, openSUSE 15.3+. Other distros require glibc 2.28+ (ldd --version to check)
  • .AppImage should work everywhere
  • Make sure you have latest graphics drivers installed
  • Possibly needed packages: sudo apt install libva2 libvdpau1 libasound2 libxkbcommon0 libpulse0 libc++-dev vdpau-va-driver libvulkan1
  • GPU specific packages:
    • NVIDIA: nvidia-opencl-icd nvidia-vdpau-driver nvidia-egl-icd nvidia-vulkan-icd libnvcuvid1 libnvidia-encode1
    • Intel: intel-media-va-driver i965-va-driver beignet-opencl-icd intel-opencl-icd
    • AMD: mesa-vdpau-drivers mesa-va-drivers mesa-opencl-icd libegl-mesa0 mesa-vulkan-drivers
• Android 6+

Help and support

For general support and discussion, you can find the developers and other users on the Gyroflow Discord server.

For companies or people wishing to get in touch with the team privately for collaborative purposes: [email protected].

Roadmap

See the open issues for a list of proposed features and known issues. There's also a ton of TODO comments throughout the code.

Video editor plugins

Gyroflow OpenFX plugin is available here.

Adobe After Effects and Davinci Resolve plugins are planned, but not ready yet

Contributing

Contributions are what make the open source community such an amazing place to learn, inspire, and create. Any contributors are greatly appreciated.

• If you have suggestions for adding or removing features, feel free to open an issue to discuss it.
• If you want to implement a feature, you can fork this project, implement your code and open a pull request.

Translations

Currently Gyroflow is available in:

• English (base language)
• Chinese Simplified (by DusKing1)
• Chinese Traditional (by DusKing1)
• Danish (by ElvinC)
• Finnish (by Jesse Julkunen)
• French (by KennyDorion)
• Galician (by Martín Costas)
• German (by Grommi and Nicecrash)
• Greek (by Stamatis Galiatsatos)
• Indonesian (by Aloysius Puspandono)
• Italian (by Rosario Casciello)
• Japanese (by 井上康)
• Norwegian (by MiniGod and alexagv)
• Polish (by AdrianEddy)
• Portuguese Brazilian (by KallelGaNewk)
• Portuguese (by Ricardo Pimentel)
• Russian (by Андрей Гурьянов, redstar01 and lukdut)
• Slovak (by Radovan Leitman and Eduard Petrovsky)
• Spanish (by Pelado-Mat)
• Turkish (by Metin Oktay Yılmaz)
• Ukrainian (by Artem Alexandrov)

Help us translate Gyroflow to your language! We use crowdin to manage translations and you can contribute there: https://crowdin.com/project/gyroflow

I want to contribute but I don't know Rust or QML

• The Rust book is a great way to get started with Rust: https://doc.rust-lang.org/book/
• Additional useful resources for Rust: https://quickref.me/rust and https://cheats.rs/
• For the UI stuff, there's a nice QML book by The Qt Company: https://www.qt.io/product/qt6/qml-book

Development

Used languages and technologies

Gyroflow is written in Rust, with UI written in QML. It uses Qt, ffmpeg, OpenCV and mdk-sdk external dependencies for the main program, but the core library is written in pure Rust without any external dependencies.

OpenCV usage is kept to a minimum, used only for lens calibration and optical flow (src/core/calibration/mod.rs and src/core/synchronization/opencv.rs). Core algorithms and undistortion don't use OpenCV.

GPU stuff supports DirectX, OpenGL, Metal and Vulkan thanks to Qt RHI and wgpu. For GPU processing we use OpenCL or wgpu, with highly parallelized CPU implementation as a fallback.

Code structure

1. Entire GUI is in the src/ui directory
2. src/controller.rs is a bridge between UI and core, it takes all commands from QML and calls functions in core
3. src/core contains the whole gyroflow engine and doesn't depend on Qt or ffmpeg, and OpenCV is optional
4. src/rendering contains all FFmpeg related code for rendering final video and processing for synchronization
5. src/core/gpu contains GPU implementations of the undistortion
6. src/qt_gpu contains zero-copy GPU undistortion path, using Qt RHI and GLSL compute shader, but this method is experimental and buggy for now
7. src/gyroflow.rs is the main entry point
8. mod.rs or lib.rs in each directory act as a main entry of the module (directory name is the module name and mod.rs is kind of an entry point)

Dev environment

Visual Studio Code with rust-analyzer extension.

For working with QML I recommend to use Qt Creator and load all QML files there, as it has auto-complete and syntax highlighting. The project also supports UI live reload, it's a super quick way of working with the UI. Just change live_reload = true in gyroflow.rs and it should work right away. Now every time you change any QML file, the app should reload it immediately.

Building on Windows

1. Get latest stable Rust language from: https://rustup.rs/
   • Please make sure to check the English language pack option when installing the C++ build tools from Visual Studio Installer
2. Clone the repo: git clone https://github.com/gyroflow/gyroflow.git
3. Install dependencies to the ext directory: cd gyroflow/ext
   • Qt 6.3.0 or higher: pip3 install -U pip & pip3 install aqtinstall then aqt install-qt windows desktop 6.3.0 win64_msvc2019_64 -m qtshadertools or use the official installer
   • FFmpeg 5.0: download ffmpeg 5.0 lite and unzip it to ext/ffmpeg-5.0-windows-desktop-vs2022-gpl-lite
   • vcpkg: git clone --depth 1 https://github.com/Microsoft/vcpkg.git & .\vcpkg\bootstrap-vcpkg.bat -disableMetrics
   • OpenCV: .\vcpkg\vcpkg install "opencv[core]:x64-windows-release"
   • OpenCL: .\vcpkg\vcpkg install "opencl:x64-windows-release"
   • LLVM: download and install LLVM
4. Make sure you have 7z in PATH.
5. Setup the environment in powershell (or set the same variables in cmd): ./__env.ps1 - I do this in VS Code built-in terminal
   • some machines might have issue that scripts are forbidden to run, try to set-executionpolicy remotesigned in powershell with admin
6. Compile and run: cargo run --release

Building on MacOS

1. Get latest stable Rust language from: https://rustup.rs/
2. Install Xcode command line tools: xcode-select --install
3. Clone the repo: git clone https://github.com/gyroflow/gyroflow.git
4. Install dependencies: cd gyroflow/ext && ./install-deps-mac.sh
5. For building on Apple M1, change ffmpeg-x64_64 -> ffmpeg-arm64 and x64-osx-release -> arm64-osx in __env-macos.sh
6. Setup the environment in terminal: ./__env-macos.sh - I do this in VS Code built-in terminal
7. For some reason DYLD_FALLBACK_LIBRARY_PATH may be overwritten so export it again in terminal: export DYLD_FALLBACK_LIBRARY_PATH="$(xcode-select --print-path)/Toolchains/XcodeDefault.xctoolchain/usr/lib/" (if you have build errors, try the other one from __env-macos.sh)
8. Compile and run: cargo run --release
9. If it fails to run, do: ./_deployment/deploy-macos.sh once

Building on Linux

1. Get latest stable Rust language from: https://rustup.rs/
2. Clone the repo: git clone https://github.com/gyroflow/gyroflow.git
3. Install dependencies: cd gyroflow/ext && ./install-deps-linux.sh (Debian based apt)
4. Setup the environment in terminal: ./__env-linux.sh - I do this in VS Code built-in terminal
5. Compile and run: cargo run --release

Building for Android

1. Android is not well supported yet, but the app can be built and somewhat works. So far only building on Windows was tested
2. Install Qt for Android: aqt install-qt windows android 6.3.0 android_arm64_v8a and aqt install-qt windows desktop 6.3.0 win64_mingw
3. Install cargo-apk: cargo install --git https://github.com/rust-windowing/android-ndk-rs.git cargo-apk
4. Add a Rust target: rustup target add aarch64-linux-android
5. Update Cargo.toml to comment out [[[bin]] section and uncomment [lib] section
6. Patch C:\Users\you\.cargo\registry\src\github.com-1ecc6299db9ec823\opencv-0.61.3\build.rs: Change if cfg!(target_env = "msvc") to if std::env::var("CARGO_CFG_TARGET_ENV").unwrap() == "msvc"
7. Update paths in _deployment/build-android.ps1
8. Run .\_deployment\build-android.ps1 in Powershell

Profiling on Windows

1. Install and run Visual Studio Community Edition
2. Compile and run GyroFlow with the profile profile: cargo run --profile profile
3. In Visual Studio, go to Debug -> Performance Profiler...
   • Under Target, open Change Target and select Running Process..., select the running gyroflow.exe process

License

Distributed under the GPLv3 License. See LICENSE for more information.

Authors

• AdrianEddy - Author of the Rust implementation (code in this repository), author of the UI, GPU processing, rolling shutter correction and advanced rendering features
• Elvin Chen - Author of the first version in Python, laid the groundwork to make all this possible

Notable contributors

• Maik Menz - Contributed to all areas of Gyroflow with fixes and improvements
• Aphobius - Author of the velocity dampened smoothing algorithm
• Marc Roeschlin - Author of the adaptive zoom algorithm
• Ilya Epifanov - Author of the OpenFX plugin

Acknowledgements

• Gyroflow Python version (legacy code)
• telemetry-parser