RapidRAW
GPU & Drivers
RapidRAW renders every edit on the GPU using wgpu, which talks to your graphics card through DirectX 12 or Vulkan (Windows), Metal (macOS), Vulkan (Linux) or OpenGL as a fallback.
Typical GPU problems:
- The image area stays black, blank or transparent.
- The app crashes when you open a photo.
- Editing is very slow even though you have a good graphics card.
1. Update your graphics driver
Get the latest driver from NVIDIA, AMD or Intel directly. Drivers installed by Windows Update are often old. On Linux, use your distribution's current Mesa, or the current NVIDIA driver.
2. Change the Processing Backend
Open Settings → Processing → Processing Backend:
| Option | Platforms | Notes |
|---|---|---|
| Auto (default) | All | wgpu picks the best one. |
| Vulkan | Windows, Linux | Fast, usually the best on Linux. |
| DirectX 12 | Windows | Usually the best on Windows. |
| Metal | macOS | The only real choice on Mac. |
| OpenGL | All | Slower, but works almost everywhere. Use it when nothing else works. |
Click Save & Relaunch. Try the options in this order: Windows DirectX 12 → Vulkan → OpenGL, macOS Metal → OpenGL, Linux Vulkan → OpenGL. (The list may also show DirectX 12 on Linux, but it does not work there.)
If RapidRAW crashed while starting the GPU on the previous launch, it switches to OpenGL by itself the next time so you can get back into the app.
3. Try without WGPU Direct Rendering (Windows, macOS)
Settings → Processing → WGPU Direct Rendering draws the preview straight into the window. It is much faster, but on a few systems (some older Intel graphics, unusual multi-monitor setups) the image stays transparent or the interface disappears. Turn it off, click Save & Relaunch, and see if the problem is gone.
Symptoms
| Symptom | Likely cause | Try |
|---|---|---|
| Image area transparent or black | Problem creating the GPU surface | Turn off WGPU Direct Rendering, then try another backend. |
| Crash when opening a photo | Driver issue with the selected backend | Change the backend (above). |
| Sliders lag a lot | GPU too slow for the preview size, or running on the integrated GPU | Lower Render Resolution Scale or Live Preview Quality in Settings, and force the dedicated GPU (below). |
| Very slow despite a fast GPU | Running on a software renderer | Check app.log for "llvmpipe" or "Microsoft Basic Render Driver"; install a proper driver. |
| Flickering when zooming | Driver bug | Update the driver. |
The log records which GPU and backend RapidRAW uses. See Filing a Bug Report for where to find it.
Use the dedicated GPU
Laptops with integrated and dedicated graphics sometimes run RapidRAW on the slower integrated chip.
Windows
Settings → System → Display → Graphics, add RapidRAW.exe, then Options → High performance. NVIDIA users can also set it in the NVIDIA Control Panel under Manage 3D settings → Program Settings.
macOS
Apple Silicon Macs have one GPU, nothing to do. On Intel MacBook Pros with two GPUs, turn off Automatic graphics switching in System Settings → Battery while editing.
Linux
# AMD/Intel hybrid laptops
DRI_PRIME=1 RapidRAW
# NVIDIA Optimus
prime-run RapidRAW
Replace RapidRAW with ./RapidRAW_*.AppImage or flatpak run io.github.CyberTimon.RapidRAW depending on how you installed it.
AI features: CPU or GPU?
AI features run through ONNX Runtime, separately from the main editor:
| Platform | AI on the GPU | AI on the CPU |
|---|---|---|
| Windows (DirectML) | Sky, Foreground, Text and Face detection, Depth, AI Denoise, tagging, normal maps | Subject masks, face parts, Quick Eraser |
| macOS (Core ML) | Depth, AI Denoise | Everything else |
| Linux, Android | none | All |
GPU acceleration is controlled by Settings → General → Generative AI → Built-in → GPU Acceleration (Windows and macOS, on by default). If AI masks or denoise crash, hang or give strange results, turn it off. RapidRAW also falls back to the CPU on its own if the GPU fails.
A faster CPU makes CPU-based AI features faster. The rest of the editor (all adjustments, masks blending, effects) always runs on the GPU in real time.
Still not working?
Open an issue with the Bug Report template and include the log, your GPU model and driver version, and which backends you tried. See Filing a Bug Report.