models


El Michael XL Vpred v1

-Fine tune Model
-Base model:NoobAI XL V-Pred v1.0
-dmd2_sdxl_4step_lora included

-(forge)turn on Zero Terminal SNR
***Turn off when using other epred models


Download: v1

- base model

-NoobAI XL V-Pred v1.0 -great XL model!

- trigger

illust style, (or/and: digital illustration,)
anime style, (or/and: high-quality anime drawing,)
simple illust style, solid circle eyes, (or/and: flat color,)
manga style, monochrome,
pc98 style, dithering, (or/and: jaggy lines,)

- quality tags

masterpiece, best quality, 

- negative tags

low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic

- prompt

danbooru tag + (natural english: *not good)

- Sampling method

Euler a :6-16 steps

dmd2 included: dmd2_sdxl_4step_lora

- Schedule type

SGM Uniform, Normal, Simple 
(Low saturation) KL Optimal

- cfg scale

1-1.5 (1.07 is good for me.)

- Resolution

768 x1024, 1024 x 768
1024 x 1024
896 x 1152, 1152 x 896
512 x 768, 768 x 512

sample prompt

- illust style

masterpiece, best quality,
illust style, 1girl, solo, angel wing, halo, blonde mendium hair, blue eyes, white dress, smile, dark sky, romantic setting
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 6, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 1.3, Seed: 1830545276, Size: 896x1152, Model hash: 77c1f90075, Model: El_Michael_XL_Vpred_v1, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 1.3, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae


- anime style

masterpiece, best quality,
anime style, 2girls, school uniform, smile, classroom, windw, from side, sunset, from below, face to face,
Left girl has a black hair with long hair and green eyes ,
Right girl has a brown hair with ponytail and yellow eyes
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 6, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 1.3, Seed: 2747030854, Size: 896x1152, Model hash: 77c1f90075, Model: El_Michael_XL_Vpred_v1, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 1.3, Hires upscale: 2, Hires upscaler: 4x-UltraSharp, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae


- simple illust style

masterpiece, best quality,
simple illust style, flat color, solid circle eyes, 1girl, solo, red long hair, brown pajama \(kigurumi\), white back ground, holding pillow, on, bed, looking at viewer
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 6, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 1.3, Seed: 877723997, Size: 896x1152, Model hash: 77c1f90075, Model: El_Michael_XL_Vpred_v1, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 1.3, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Emphasis: No norm, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae


- manga style *Oops, the dialogue won't go away.

masterpiece, best quality,
manga style, monochrome, 1girl, solo, angry, twintails, school uniform, motion lines, shop
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 6, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 1.3, Seed: 3282550249, Size: 896x1152, Model hash: 77c1f90075, Model: El_Michael_XL_Vpred_v1, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 1.3, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae


- pc98 style

masterpiece, best quality,
pc98 style, dithering, 1girl, solo, blue hair, purple eyes, confuse, gothic dress, holding candle, dark castle
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 6, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 1.3, Seed: 3942053457, Size: 896x1152, Model hash: 77c1f90075, Model: El_Michael_XL_Vpred_v1, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 1.3, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae


no dmd2

*I don't know what a good setting is. (Download from "Files" at the top of the page)

masterpiece, best quality,
illust style, 1girl, solo, angel wing, halo, blonde mendium hair, blue eyes, white dress, smile, dark sky, romantic setting
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 30, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 7, Seed: 3698065647, Size: 896x1152, Model hash: 7a9347dd09, Model: El_Michael_XL_Vpred_v1_no_dmd2, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 7, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae

masterpiece, best quality,
anime style, 1girl, solo, angel wing, halo, blonde mendium hair, blue eyes, white dress, smile, dark sky, romantic setting
Negative prompt: low quality, worst quality, artist name, signature, watermark, lowres, bad anatomy, text, speech bubble, comic
Steps: 30, Sampler: Euler a, Schedule type: SGM Uniform, CFG scale: 5, Seed: 683797657, Size: 896x1152, Model hash: 7a9347dd09, Model: El_Michael_XL_Vpred_v1_no_dmd2, Denoising strength: 0.3, Hires Module 1: Use same choices, Hires CFG Scale: 5, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Noise Schedule: Zero Terminal SNR, Version: f2.0.1v1.10.1-previous-634-g37301b22, Module 1: sdxl.vae



-Train Settings

base model: NoobAI-XL-Vpred-v1.0.safetensors

caption: JoyCaption Alpha Two
tags: WD EVA02-Large Tagger v3

    --network_module "sdxl_train.py" ^
    --caption_dropout_rate="0" ^
    --vae_batch_size="1" ^
    --gradient_checkpointing ^
    --persistent_data_loader_workers ^
    --cache_latents ^
    --cache_latents_to_disk ^
    --max_data_loader_n_workers=2 ^
    --enable_bucket ^
    --bucket_no_upscale ^
    --save_model_as "safetensors" ^
    --mixed_precision "bf16" ^
    --learning_rate=5e-6 ^
    --train_text_encoder ^
    --learning_rate_te1=1e-6 ^
    --learning_rate_te2=1e-7 ^
    --resolution=1024,1024 ^
    --train_batch_size 2 ^
    --optimizer_type "adafactor" ^
    --optimizer_args "relative_step=False" "scale_parameter=False" "warmup_init=False"  ^
    --lr_scheduler "constant_with_warmup" ^
    --save_precision "bf16" ^
    --caption_extension ".txt" ^
    --no_half_vae ^
    --full_bf16 ^
    --max_grad_norm=0 ^
    --min_snr_gamma=5 ^
    --max_token_length=225 ^
    --fused_backward_pass ^
    --v_parameterization ^
    --zero_terminal_snr ^
    --save_state ^
    --xformers

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Examples
Unable to determine this model's library. Check the docs .

Model tree for deadman44/SDXL_Anime_Merged_Models

Finetunes
2 models