Gaz's Hedcut LoRa版本v1.1 (ID: 33513)

Hedcuts refer to the style of manually traced halftoning applied to photographs to capture their content and essence in an elegant way by use of stippling and hash techniques, fit for small form printing.

A similar aesthetic to the product of woodblock printing.

The style is most associated with the portraiture of the Wall Street Journal, begun by illustrator Kevin Sprouls, wherein they use the hedcut technique for authors or subjects of articles to accompany them as graphic representation elements in lieu of real photographs.

This LoRa was trained in Realistic Vision 1.4 at 64/64 net dim and tends to produce solid results weighted at 1 without invocation, though if used in another model or if it starts slipping you can reign it back by suffixing your prompt with "in hedcut style"

Best practices are to conjure portraits of people or things and if need be use the active tag with short to the point prompts, if it's still losing grip on the aesthetic dial up the weight to 1.1+

Works pretty good with ControlNet anchoring as well to apply the style to real photos

描述:

Changed up tags and pruned data a bit then trained for more epochs, I can't objectively say which is better so it's not a replacement, but this one seems stronger in general and inside other models so it's a bit of an alternative perhaps.
Doesnt need trigger word but "in hedcut style" will steer it back if it starts slipping.

训练词语: in hedcut style