What are the best models avaliable for doing portraits & abnormal characters?

[BenneyR]

So I'm really new to this.

I am not that great with tech, but as an artist, I really think AI can help a ton with reference material, so that's why I'm here.

I understand that having a good resource of models can help you get better results and I wanted to know which models are the best currently for making good portraits.

I also am looking for models that help with abnormal characters like aliens, monsters, furries, and anything in this strange and cool realm.

Also, what negative prompts actually give you better results?
I understand they are meant to improve your batches/results but what terms have been found to actually be the most effective?

I hope that I'm clear and I really appreciate any help that you might give!

[drax-xard]

I've had varying results with negative prompts, if you go overboard the model will flip the f out and you start getting worst results.
I usually have them separated in categories and add/remove them accordingly, something like this:

Defects:
(extra fingers), (mutation), (mutated), (mutant), (deformed), (disfigured), (mutilated), (blurry), (extra limbs), (glitchy), (extra breasts), (double torso), (extra legs), (extra arms), (extra hands), (mangled fingers), (missing breasts), (missing lips), (ugly face), (conjoined), (cleft lips)

Text and graphics:
text, caption, ads, cover, logo, sign, icon, label, meme, quora, vector, symbol, inspirational, books, comics, phrases, graphics, anime, thumbnail, wallpaper, abstract

SEO crap:
royals, royalty, trump, biden, clinton, putin, wtf, buy, sale, sell, seller, product, premium, comics, fashion, perfume, marines, news, samsung, iphone, credit, usb, bluetooth, wireless, mp3, coronavirus, covid, virus, vaccine, baseball, traduction, basketball, football

Brands and websites:
facebook, ebay, bestbuy, costco, craigslist, google, minecraft, samsung, iphone, alibaba, wallmart, rakuten, livescore, ikea, bet365, gmail, olx, discord, twitter

Depending on the composition of your desired image you should tweak them untill you get a set that works for the "theme" you're currently going for.

As for positives, it's fickle black magic. Changing the order, arranging them by [subject], [characteristics], [verb/scenery] sometimes help, sometimes it doesn't.

And as for models (as in the models used by the software), the default works best for general purpose, but there are a number of genre/style specific poppin up pretty much every day.

You can find a list of models here: https://rentry.org/sdmodels
Be careful tho, don't install random models without checking their source.