modelle/EthernalDope - v1.0

EthernalDope - v1.0

|
10/23/2023
|
4:15:12 PM
| Discussion

Verwandte Schlüsselwörter & Tags

Empfohlene Prompts

masterpiece,best quality,ultra-detailed

best quality,solo,1girl,looking at viewer

Empfohlene Negative Prompts

EasyNegativeV2,bad-hands-5,blurry,bad anatomy,long legs,poorly detailed,sketch,multiple views,worst quality, low quality:1.4

Empfohlene Parameter

samplers

DPM++ 2M Karras

steps

20 - 25

cfg

7

clip skip

2

resolution

512x768

vae

vae-ft-mse-840000-ema-pruned.safetensors

other models

makima_offset, Tohru, gotoh_hitori, anime, JinxLol, amber_genshin, Fiolina

Empfohlene Hires (Hochauflösungs-) Parameter

upscaler

R-ESRGAN 4x+ Anime6B, 4x-AnimeSharp

upscale

2

steps

10 - 13

denoising strength

0.4

Tipps

The generation of hands can be somewhat bad 😿, so it is recommended to use a ControlNet or Lycoris model.

Despite being a merge of models that offer 3D results, the model's output can vary to 2D depending on the lora used.

Ersteller-Sponsoren

Try now EthernalDope on Shakker!!! (and new models that I will upload there first)

Recommended settings

Prompt: (masterpiece, best quality, ultra-detailed)

Negative Prompt: EasyNegativeV2, bad-hands-5, blurry, bad anatomy, long legs,

poorly detailed, sketch, multiple views, (worst quality, low quality:1.4)

Sampler: DPM++ 2M Karras

Sampling Steps: 20 -25

Upscaler: 4x-AnimeSharp / R-ESRGAN 4x+ Anime6B

Hires steps: 10-18

Denoising Strength: 0,4-0,6

Notes:

  • The generation of hands can be somewhat bad 😿, so it is recommended to use a ControlNet or Lycoris model

  • Despite being a merge of models that offer 3D results, the results that this model can generate will vary depending on the lora that is used, giving results in 2D.

Zurück
Anime Illust Diffusion XL - v0.5-alpha
Weiter
DevlishPhotoRealism - v4.0

Modell-Details

Modelltyp

Checkpoint

Basismodell

SD 1.5

Modellversion

v1.0

Modell-Hash

45450b551b

Ersteller

Diskussion

Bitte log in um einen Kommentar zu hinterlassen.