Upscayl Guides and Troubleshooting

First Upscale Walkthrough

1. Install Upscayl from the download page and confirm your SHA256 hash matches the published value for your version.
2. Confirm you have a Vulkan-compatible dedicated GPU with up-to-date graphics drivers.
3. Launch Upscayl and open the image picker. Select a PNG, JPEG, or other supported source file.
4. Choose an AI model suited to your content (photo vs illustration). Preview if available.
5. Set your scale factor (2x, 4x, or higher depending on VRAM and quality needs).
6. Start upscaling and wait for GPU processing to complete. Duration scales with resolution and model.
7. Review the output comparison, then save to your configured output folder.

GPU and Vulkan Troubleshooting

Problem: Upscayl does not start or fails immediately

Upscayl does not work without a GPU. CPU and iGPU-only systems are unsupported. Verify you have a discrete GPU with Vulkan support.

Solution steps

1. Update NVIDIA, AMD, or Intel Arc drivers to the latest stable release for your OS.
2. On Linux, install Mesa/Vulkan packages for your distribution (vulkan-utils, mesa-vulkan-drivers).
3. On laptops, force the discrete GPU for Upscayl in Windows Graphics Settings or NVIDIA Control Panel.
4. Restart the application after driver changes. Reboot if Vulkan ICDs were newly installed.

Problem: Out of memory or very slow processing

Reduce scale factor, process smaller crops, or close GPU-heavy applications. Double Upscayl multiplies time and memory pressure; use only when needed.

Antivirus Warnings Explained

Security software may flag lesser-known executables that pack native GPU libraries. Upscayl is open source and widely mirrored; false positives are reported occasionally in community channels.

Recommended safe workflow

1. Download only from official distribution points configured on our download page.
2. Compute SHA256 and compare to the published checksum on download.html.
3. If hash matches and antivirus still blocks, submit the file to your vendor as a false positive.
4. Add a folder exclusion only after hash verification, not before.
5. Never disable antivirus entirely on internet-connected machines.

Batch Processing Best Practices

Organize source images in a dedicated input folder. Use consistent model and scale settings across a batch for uniform deliverables.

Process overnight for large archives. Monitor GPU thermals on sustained jobs. Start with a small test subset before running thousands of files.

Keep originals immutable. Write outputs to a separate directory with clear naming so you can re-run with different models without data loss.

Choosing the Right AI Model

Different models emphasize sharp edges, skin texture, flat illustration areas, or anime line art differently. If results look oversharpened or plastic, switch models before raising scale.

• Photographs: start with general Real-ESRGAN variants.
• Digital art and UI assets: try illustration-friendly models.
• Anime and line art: use specialized anime models to avoid muddy gradients.
• Heavily compressed JPEGs: consider Double Upscayl after a satisfactory single pass.

Output Settings and Quality

Match output format to destination: PNG for lossless intermediates, JPEG for web delivery with tuned compression. Extremely high scale factors produce very large files; plan storage accordingly.

For print workflows, prefer PNG outputs and verify DPI metadata in your layout tool after export. Color accuracy is best preserved when starting from highest-quality sources available.