modelos/Ketchup-PDXL - final

Ketchup-PDXL - final

10/14/2025

9:17:04 AM

Palabras Clave y Etiquetas Relacionadas

animagine xl,anime,anime style,modelo base,checkpoint,final,girls,ketchup-pdxl,model merge,neverwinter13,omega pony xl,pony,ponyxl,snowpony,stable diffusion model,mujer

A mature female with thick curly black hair, greenish hazel eyes, wearing a purple velvet gothic lolita dress leans forward beside a spa pool within classical stone pillars.

Voluptuous woman with curly black hair, greenish hazel eyes, and a surprised expression stands in an enchanted forest surrounded by red-capped mushrooms wearing a purple gothic lolita dress.

Voluptuous anime-style female with curly black hair and greenish hazel eyes wearing a gothic lolita attire crouching in an outdoor alley with platform boots and lace stockings.

Prompts Recomendados

score_9,score_8_up,score_7_up,masterpiece,best quality,very aesthetic,absurdres

score_9, score_8_up, score_7_up, masterpiece, best quality, very aesthetic, absurdres, 1girl, red eyes, blurry, blurry background, blurry foreground, bokeh, depth of field, lens flare, light particles

Prompts Negativos Recomendados

score_6,score_5,score_4,score_3,score_2,score_1,ugly,bad hand,bad feet,extra feet,lowres,(bad),text,error,fewer,extra,missing,worst quality,jpeg artifacts,low quality,watermark,unfinished,displeasing,oldest,early,chromatic aberration,signature,extra digits,artistic error,username,scan,abstract,sfw,extra digits,mutated hands

score_6, score_5, score_4, score_3, score_2, score_1, ugly, bad hand, bad feet, extra feet, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, artistic error, username, extra digits, scan, abstract, sfw, mutated hands, nipples

Parámetros Recomendados

samplers

Euler Ancestral, Euler A, Euler a

steps

20 - 26

cfg

6 - 7

resolution

512x768, 768x512, 768x768

Parámetros Recomendados de Alta Resolución

upscaler

4x Ultra Sharp, 4x-UltraSharp

upscale

2 - 4

steps

10 - 15

denoising strength

0.3

Consejos

Use lower classifier-free guidance scale to improve quality, around 6-7.

Sampling steps between 20 and 26 are recommended for best results.

Euler Ancestral (Euler A) sampler provides better image quality.

Aspectos Destacados de la Versión

Initial/final merge.

Used TelePony-v3-experimental's block weights for Alpha and TelePony-v3-final for Beta.
Great with backgrounds/scenery. Lora support is also good.
NSFW Capabilities are great. Some are lost but can still be generated with 'more' prompting.
Some hiccups when it comes to F ups like distorted anatomy and such. Change resolution if needed.
VAE is baked in

Model Details.

Initial/Final merge.

r_ain for SnowPony.
Minthybasis for 4th tail.

Big thanks to Goofy_Ai, r_ain, Minthybasis, kitarz, WAI0731, CagliostroLab, and the entirety of Civitai and the Generative AI Community for giving such inspiration, shits and giggles. Helped me out a lot in dealing with my mental health issues. You guys are the best.

Recommended settings.

it's recommended to use a lower classifier-free guidance (CFG Scale) of around 6 - 7, sampling steps between 20 and 26, and to use Euler Ancestral(Euler A) as a sampler. For upscaler, I went with 4x Ultra Sharp

Suggested prompts: "score_9, score_8_up, score_7_up, masterpiece, best quality, very aesthetic, absurdres,"

Suggested negative prompts: "score_6, score_5, score_4, score_3, score_2, score_1, ugly, bad hand, bad feet, extra feet, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, abstract, sfw, extra digits, mutated hands"

Notes

Based on SnowPony and 4th tail, Omega Pony XL, WAI-ANI-NSFW-PONYXL and Animagine XL. Gone with Merge Block Weighted option with Sum Twice and underwent selective block weight test, more like trial and error or collect, select then retain. All credits goes to the original authors of the models and resources used for this merge. This could be the initial and final version of this, as I'm planning to quit or go on an indefinite hiatus.

Following Fair AI Public License 1.0-SD license.