Large dimensions can increase repeated subjects or distorted anatomy
An image much larger than the model's usual training resolution can contain repeated subjects, distorted anatomy, or inconsistent composition. The outcome depends on the model, prompt, and dimensions.Two options can help:
generates a smaller image, then enlarges and refines it. The refinement can change details.
An upscaler enlarges the finished image. It can sharpen or invent details while usually retaining the overall layout.
Compare the result with the original at full size. Neither method guarantees that anatomy or other errors will be corrected.