Model reference · open weights

WAI-REAL_CN

Image TheImposterImposters Text→image 1 build Licence not stated 1k dl/mo

WAI-REAL_CN is an open-weight image model from TheImposterImposters. WAI-REAL_CN-v11.0 (FP32) weighs 6.9 GB; the smallest configuration that runs it is RTX 3060 12 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

  • WAI-REAL_CN is a text-to-image model with 2.6B parameters, developed by TheImposterImposters and based on the Pony architecture.
  • The v11.0 version focuses on optimizing light, shadow, texture, and lens performance through photography-specific training.
  • The model card provides recommended settings, including a CFG scale of 5 to 7 and the Euler a sampler, along with specific positive and negative prompts for generation.

Summary of the TheImposterImposters/WAI-REAL_CN-v11.0 model card, 2026-10-01

What it is

Released byTheImposterImposters
Released2024-09-12
Parameters2.6B
VRAM6.9 GB for the weights

What it runs on

Memory and cards for WAI-REAL_CN-v11.0 (FP32)

6.9 GBweights, file size
5.1 GBbiggest part
5.0 GBworking memory, one 1024² image
537 MBruntime overhead
CardOne imageMemory
RTX 3060 12 GBtight (encoders offloaded)11.6 GB
RTX 4060 Ti 16 GBfits15.4 GB
RTX 3090 24 GBfits23.4 GB
RTX 4090 24 GBfits23.4 GB
RTX 5090 32 GBfits31.0 GB
L40S 48 GBfits44.0 GB
A100 80 GBfits78.2 GB
H100 80 GBfits78.1 GB
RTX PRO 6000 Blackwell 96 GBfits93.8 GB
DGX Spark (GB10) 128 GB unifiedfits107 GB
H200 141 GBfits138 GB
B200 180 GBfits176 GB

From the model card

What TheImposterImposters says about WAI-REAL_CN

Read the model card
  • ID: 469902
  • Author: WAI0731
  • Type: Checkpoint
  • Base Model: Pony
  • Base Model Type: Standard
  • Model Version ID: 782802

关于buzz

由于FLUX模型的出色表现,我也将对FLUX的大模型进行训练,但是由于硬件的原因我只能租GPU来实现训练,所以可能会需要使用很多现金.所以我最近更新的模型将会进行一定的BUZZ回收来填补自己的资金.

Due to the excellent performance of the FLUX model, I will also train my checkpoint based on FLUX. However, due to hardware limitations, I can only rent GPUs for training, which may require a significant amount of cash. Therefore, my recent model updates will involve some BUZZ collection to cover my expenses.

About V11

对v10版本整体权重调整以及进行了一点摄影方面得训练跟调整.

在光影,纹理,镜头表现方面有一定得优化.

The user has made overall weight adjustments to the v10 version and has also conducted some training and adjustments related to photography.

The user has made optimizations in light and shadow, texture, and lens performance.

V11

Recommended Setting

  • Steps: 30
  • CFG scale: 5~7
  • Sampler: Euler a
  • ADetailer face_yolov8s.pt use (ADetailer sampler: DPM++ 2M, ADetailer scheduler: Karras )

Positive Prompt

score_9, score_8_up, score_7_up,source_real,real skin,
BREAK

Negative Prompt

worst quality,low quality,
worst detail,low detail,
source_furry, source_pony, source_cartoon,2d,
,bad anatomy, bad hands, text, error, missing fingers, (extra digit:1.4), fewer digits, cropped, bad artist, outlines,jpeg artifacts, signature, watermark, username, blurry,3d,

__________________________________________________________________________________________________

V10 简单使用说明

V10 Simple Usage Guide

1.对于只想简单生成图片的用户,请你务必按照我的推荐设置来使用,尤其是负面,请不要随意改动.CFG请只使用7.

1.For users who just want to generate images simply, please make sure to follow my recommended settings, especially for negative prompts. Please do not change them arbitrarily. CFG just use 7.

2. 对于上级用户,我的推荐是删除正面提示里面的"realistic,photorealistic, ",然后请从cfg3.5开始生成如果效果不满意再请增加cfg最大到7.如果结果还是不满意,在考虑增加写实tag.

2.For advanced users, my recommendation is to delete Positive Prompt"realistic,photorealistic, " and start generating with a CFG scale of 3.5. If the result is not satisfactory, please increase the CFG gradually, up to a maximum of 7.If the results are still unsatisfactory, consider adding realistic tag.

3.关于负面提示,请不要随意删除,这会影响生成图像效果.当然对于上级者来说请随意编辑,如果有更好的提示请务必给我留言.

3.Regarding negative prompts, please do not delete them arbitrarily, as this can affect the image generation quality. However, advanced users are welcome to edit them freely. If you have better prompts, please make sure to leave me a message.

V10

Recommended Setting

  • Steps: 30
  • CFG scale: 3.5~7
  • Sampler: Euler a
  • ADetailer face_yolov8n/s.pt use (ADetailer sampler: DPM++ 2M, ADetailer scheduler: Karras )

Positive Prompt

score_9, score_8_up, score_7_up,source_real,realistic,photorealistic,
BREAK

Negative Prompt

worst quality,low quality,
worst detail,low detail,
source_furry, source_pony, source_cartoon,2d,
,bad anatomy, bad hands, text, error, missing fingers, (extra digit:1.4), fewer digits, cropped, bad artist, outlines,jpeg artifacts, signature, watermark, username, blurry,3d,

__________________________________________________________________________________________________

V2toV9

Recommended Setting

  • Steps: 30-50
  • CFG scale: 7
  • Sampler: Euler a
  • Sampler: Euler a ADetailer face_yolov8n/s.pt use more better

Positive Prompt

score_9, score_8_up, score_7_up,

Negative Prompt

score_6,score_5,score_4, worst quality, low quality, bad anatomy, bad hands, missing fingers, fewer digits, source_furry, source_pony, source_cartoon,3d, blurry,

________________________________________________________________________________________________

v1

Recommended Setting

  • Steps: 30
  • CFG scale: 3.5-7
  • Sampler: Euler a or DPM++ 2M Karras
  • ADetailer face_yolov8n/s.pt use more better

Positive Prompt

score_9, score_8_up, score_7_up,

Negative Prompt

score_6,score_5,score_4,bad anatomy, bad hands, missing fingers, fewer digits, source_furry, source_pony, source_cartoon,3d, blurry,


Content copied from Civitai: WAI-REAL_CN on 2024-09-12 08:42:23.523792+00:00

Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.

How it works

How image models work

Text promptwhat to makeText encoderunderstands itDiffusion stepsdenoise to pixelsImagePNG / JPEGA diffusion model starts from noise and denoises it, guided by your prompt, into a finished image.

Running it yourself

Run it on a rented GPU

Rent a machine by the hour. ComfyUI is installed on it. Open ComfyUI through the tunnel: its default workflow loads a checkpoint. Choose this model's file in Load Checkpoint, with the settings its model card gives.

# on your rented machine: pip install diffusers transformers accelerate
import torch
from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained("TheImposterImposters/WAI-REAL_CN-v11.0", torch_dtype=torch.bfloat16).to("cuda")
image = pipe(prompt="a red bicycle on a cobbled street").images[0]
image.save("/workspace/out.png")
Renting a GPU: connect, tunnels, Python
# on your rented machine (the ssh line is on its page in the console)
# get REPO FILE FOLDER: one file into /workspace/models/FOLDER, where ComfyUI loads it from
get() { hf download "$1" "$2" --local-dir /workspace/hf-files && mkdir -p "/workspace/models/$3" && mv "/workspace/hf-files/$2" "/workspace/models/$3/$4"; }

# the model (6.5 GB)
get TheImposterImposters/WAI-REAL_CN-v11.0 waiREALCN_v110.safetensors checkpoints

start-comfyui
Renting a GPU: connect, tunnels, ComfyUI
# on your computer, in a second terminal: ComfyUI in your browser at http://localhost:8188
# HOST, PORT and BOOKING are your machine's, from its page in the console
ssh -o HostKeyAlias=axforge-BOOKING -o StrictHostKeyChecking=accept-new -L 8188:localhost:8188 dev@HOST -p PORT
© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms