Small pictures restored with and without the LoRA

In each picture: the small input, the result without the LoRA (amber) and with it (teal), and a marked section enlarged underneath. Same workflow, prompt and seed. Without the LoRA is the same workflow with the Restore row switched off.

On 37 small pictures it never trained on, the result sat about 0.3 px off the full-size picture with the LoRA and more than 2 px without. Blurry photo? Lower the LoRA to 0.5.

The workflow is inside every still: drop one on ComfyUI. The full numbers are on the Hugging Face model card.