Model dedicated to my fav kpop gg TWICE trained based on dreamlike-diffusion v1.0 model.

Suggested prompt format for best results (remove [ ] when actually prompting):

[trigger word], [describe place, clothes, action, etc], beautiful, gorgeous big anime eyes, illustration, realism, highly detailed, dreamlikeart, [artist 1], [artist 2], [artist 3]

Artists are optional but I suggest artist 1 as Jeremy Geddes and artist 2 is Warren Louw. Artist 3 can be random for variety. The first 2 artist sets a better base. Please feel free to experiment and share your outcome.

Suggested negative prompts (just copy paste this and adjust where necessary - for some reason it improves the output quality rather than actually doing all those negatives lol):

low-res, ugly, bad anatomy, extra limbs, missing limbs, grotesque, malformed, deformed, asymmetrical eyes, text, words, signature, watermark

Suggested samplers are Euler, Euler A, and DDIM. I personally LOVE DDIM.

Steps between 30-50 should be good.

CFG between 7 to 12 is good.

Best resolution is similar to dreamlike-diffusion:

512 x 768 (2:3)

768 x 512 (3:2)

640 x 640 (1:1)

As usual, feel free to experiment and share your results here.

Thank you for your time and interest.

Search AI Models & Addons

very few results