Hires fix
Hires fix (high-resolution fix) is a two-stage technique: generate at a safe native resolution, then upscale and run a second, partial diffusion pass so the larger image gains real detail without the artifacts of generating big in one go.
Generating directly at a very large resolution tends to break a diffusion model - you get duplicated heads, repeated limbs and other "outside its training" artifacts. Hires fix is the standard workaround. It splits the job in two so you get both a coherent composition and a high-resolution result.
How it works
- Stage 1: generate at the model's comfortable native resolution, so the overall composition is clean.
- Stage 2: upscale that image, then run a second denoising pass at the larger size with a low-to-medium denoise strength.
The key control is the denoise strength on the second pass. Low values (around 0.3-0.5) add crisp detail while preserving the original composition; push it too high and the upscale pass starts inventing new content and drifting from your first image. It pairs with plain upscaling, which enlarges without a second diffusion pass.
Why it matters
Hires fix is how you get large images that are genuinely detailed, not just stretched. It is the difference between a 4K render with real texture and a soft, upsized thumbnail - essential for posters, wallpapers and anything printed.
Try it in the generator
Put hires fix to work right now - free daily generations, commercial license included.
Frequently asked questions
What is the difference between hires fix and upscaling?
Plain upscaling enlarges an image in a single pass, faithfully but without adding new structure. Hires fix enlarges and then runs a second diffusion pass, so the model actively paints in new high-frequency detail. Hires fix yields more detail; upscaling stays closer to the original.
What denoise strength should hires fix use?
Usually 0.3 to 0.5. That is enough for the second pass to sharpen and add detail while keeping the composition from your first generation. Higher values let the upscale pass deviate and can change the image.
Related terms
- UpscalingUpscaling increases an image's resolution. AI upscalers do more than stretch pixels - they intelligently add believable detail so the larger image stays sharp.
- ResolutionResolution is the pixel size of an image - its width by height. AI models generate best at a native resolution they were trained on; to go bigger and stay sharp you generate at a sensible size and then upscale.
- DenoisingDenoising is the core operation of a diffusion model: at each step it predicts and removes a little noise, gradually turning a random field into a clear image.
- StepsSteps (sampling steps) are the number of denoising passes the model runs to turn noise into an image. More steps can mean more detail, but past a point they only add time, not quality.