Platform-Specific Issues

Platform-Specific Issues

Troubleshooting guides organized by operating system. Select your platform below.

Some issues are specific to a particular operating system. Select your platform tab below to see relevant troubleshooting information.

Gatekeeper: "Recaster Can't Be Opened"

macOS Gatekeeper blocks unsigned applications by default. Recaster builds are currently unsigned (code signing is planned for a future release).

To bypass Gatekeeper:

  1. Right-click (or Control-click) the Recaster application.
  2. Select "Open" from the context menu.
  3. Click "Open" in the confirmation dialog.
  4. This only needs to be done once. Subsequent launches work normally.

Alternatively, you can use the terminal:

xattr -d com.apple.quarantine /Applications/Recaster.app

Old rsync Version

macOS ships with rsync 2.6.9, which does not support the --info=progress2 flag used for progress reporting. This can cause errors or missing progress during file synchronization.

Fix: Install a modern rsync via Homebrew:

brew install rsync
# Verify: rsync --version should show 3.x

SwinIR Falls Back to CPU

SwinIR models cannot use CoreML acceleration on macOS because Apple's CoreML framework does not support SwinIR's dynamic input shapes. The error message mentions "shapes with dimension values of 0".

SwinIR automatically falls back to CPU processing at 2-5 FPS (vs 10-20 FPS for CoreML-accelerated Real-ESRGAN).

Recommendation

Use Real-ESRGAN for local upscaling on macOS. Quality is 9.2/10 vs SwinIR's 9.7/10 but processing speed is 4-10x faster. For SwinIR quality, use remote upscaling on a cloud GPU.

Apple Silicon (M1/M2/M3/M4) Notes

Recaster runs natively on Apple Silicon Macs. Some notes:

  • Face swapping and enhancement work at acceptable speeds via ONNX Runtime CPU provider.
  • Real-ESRGAN upscaling benefits from CoreML acceleration.
  • Local DFL training is very slow on Apple Silicon (no CUDA). Use remote training.
  • The unified memory architecture means VRAM is shared with system RAM.

Still Having Issues?

If your platform-specific issue is not covered above:

Was this page helpful?