Pepo (René Ríos Boettiger, Condorito artist)

Details

Download Files

Model description

Trained on Illustrious XL 2.0; previews genned on TewiNai with a basic workflow+face detailer

René Ríos Boettiger, better known as Pepo, brought humour to a generation of Chileans (and Argentinians and Bolivians, many of whom had no idea he was from Chile...). Granted, that humour was mainly of the Benny Hill persuasion - men are letches, women be shopping, and mothers-in-law are the devil - but hey, you try coming up with 30 pages of original jokes every month for three decades. Prior to that, he drew for slightly spicier magazines Pingüino and Can Can.

>There's already an Illustrious Pepo LORA, you even linked to it in the 'Resources' section

Unfortunately, that LORA appears to have been trained primarily on uncropped and unedited Pingüino covers. How do I know? Because every third gen plasters PINGUINO in block capitals across the top of the image. It's actually kind of impressive - I've never seen Illustrious do text so consistently! I've linked to it below if you want to compare the results for yourself, though; it's more consistent at outputting Pepo's pre-Condorito style than mine is, if nothing else.

>So what is this one trained on?

Also Pingüino covers, but I made more energetic use of the crop tool. Also plenty of Condorito pics - both full-colour covers and the limited-palette interiors - but exclusively of the human/'normie' characters.

>Wait, so no..?

Nope, Condorito himself isn't in the training data. Except for a slight fragment of his beak that I didn't edit out properly in one image, but the tagger ignored it so I assume it's fine. This here is a multi-purpose, non-specialised style LORA!

>Give me the magic words already:

·         pepostyle,: the basic style unlocker.

·         limited palette, flat colors, rust theme, white theme,: If you want to mimic the limited-palette tones of the Condorito comic interiors (see example pic #2). limited palette, orange_sky will also work in outdoors scenes, especially if you crank up the STR a little (see example pic #3). Optionally, you could also add:

·         AM halftone screening,: The printing technique used in Condorito comics interiors. Add it to your positive prompt if you want some of that texture, or stick it in negative if you want to stay nice and flat. Note that it does not play nicely with high-res fixes - when upscaling, the model will change your nice faux-analog printing texture into blatant AI digital noise. Speaking of double-edged swords:

·         old_paper_\(medium\),: Since the training data necessarily includes some ‘60s and ‘70s magazine scans, you can add this to your negative prompt if you're seeing unwelcome light brown or dull green tints. However, it's also a viable addition to your positive prompt if you want verisimilitude, and because the training data with this tag is mostly older painted magazine covers, you can also use it to encourage the model to generate proportions and shading that are closer to Pepo's older work (see example pic #4).

·         manly,: Pepo didn't make a habit of drawing attractive men - caricatures were his thing. The Chadliest-looking dude in his character collection is designated antagonist Pepe Cortisona, who at least has decent shoulders. Throwing in the 'manly' tag means the LORA is less likely to generate Quasimodo, at least (see example pic #5).

>How strong should I make the STR strength number thingy?

Note: previews were generated with TewiNai, which is very good at letting the LORA take over. If your checkpoint asserts a strong style of its own, you may need to add +0.2 to the figures below.

·         For basic cartooning layouts - medium distance, or upper body shot, side or front view, STR1.0 will give you the Pepo character designs but still allow your model's shading style and some of its visual signature (see example pic #6).

·         Because there wasn't any example art to speak of for more elaborate camera angles and tricks - from above, from below, extreme close-ups, focus on body parts - using any of those angles will mostly negate the LORA unless you crank up the STR to about 1.25. Note that at higher STR the LORA will begin to enforce flatter shading. The character proportions will also start to look a lot more cartoony (see example pic #7).

·         The maximum safe weight is around 1.45, if you're OK with the women's hourglass figures going to ridiculous extremes (use imagination, because no pics).

Images made by this model

No Images Found.