AxForge
Sign in Get started

Model reference · open weights

PonyRealism.2MainVAE

Image TheImposterImposters Text→image 1 build Licence not stated 2k dl/mo

PonyRealism.2MainVAE is an open-weight image model from TheImposterImposters. PonyRealism-v2.2MainVAE (FP32) weighs 6.9 GB; the smallest configuration that runs it is RTX 3060 12 GB.

Request Rent a GPU

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

  • PonyRealism.2MainVAE is a 2.6B parameter text-to-image model created by TheImposterImposters.
  • It is based on the Pony architecture and is designed for generating realistic images with improved lighting, color refinement, and prompt comprehension.
  • The model card recommends using Danbooru tags and specific samplers like DPM++ 2M SDE Karras for optimal results.

Summary of the TheImposterImposters/PonyRealism-v2.2MainVAE model card, 2026-10-01

Hugging Face ↗

What it is

Released byTheImposterImposters
Released2024-10-06
Parameters2.6B
VRAM6.9 GB for the weights

What it runs on

Memory and cards for PonyRealism-v2.2MainVAE (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 PonyRealism.2MainVAE

Read the model card
  • ID: 372465
  • Author: ZyloO
  • Type: Checkpoint
  • Base Model: Pony
  • Base Model Type: Standard
  • Model Version ID: 914390

✨ v2.2 IS HERE!

This update introduces small refinements across some areas, maintaining the model's current strengths while offering subtle improvements*.*

Here are the changes:

  • Improved lighting
  • Better performance in dark/night scenes
  • Improved hands and feet
  • More refined colors
  • Cleaner overall image quality
  • Slightly better face shapes from different perspectives
  • Subtle improvements to skin
  • Enhanced outputs with Euler A
  • Improved prompt comprehension
  • Slightly finer details
  • Updated VAE

🗃 Model Versions

  • v2.2: (latest) ✨

    • Main Version: Latest version with the Improvements mentioned above.
    • 4 Step Lightning Version: Soon™.
    • 8 Step Lightning Version: Soon™.

  • v2.1:

    • Main Version: Offers the highest detail but has limited compatibility with some Samplers.
    • Alternative Version: Compatible with most Samplers, though with reduced detail.
    • Inpaint Version: Refer to the "About this version" section of the version for more information.
    • 4 Step Lightning Version: Refer to the "About this version" section of the version for more information.
    • 8 Step Lightning Version: Refer to the "About this version" section of the version for more information.

  • v1.1: (old)

💡 Tips and Generation Tricks

I have written an article that explains and provides some useful resources to help you achieve great generation results. You can check it out for detailed insights and recommendations.

🏠 Recommended Parameters for Local Generation

Samplers:

  • Recommended:

    • DPM++ 2M SDE Karras
    • DPM++ 2S a Karras
    • DPM++ SDE Karras
    • DPM++ 2M SDE Exponential
    • DPM2 a
    • DPM++ 2S a
    • DPM++ 3M *
    • Euler A

Settings:

  • Steps: >=30
  • CFG: 6-7
  • Clip Skip: 2
  • Resolution: ***>***1024
  • Terminology: Recommended the use of "female/male" instead of "woman/man"

Additional Information:

  • Use Danbooru tags.

  • Don't exceed a weight of 1.5 in prompts

  • Prompt Style:

    • Positive: score_9, score_8_up, score_7_up, BREAK
    • Negative: score_4, score_5, score_6

🌎 Recommended Parameters for Site Generation

Samplers:

  • Recommended:

    • Euler a
    • DPM2 a (Best Detail)

🏋️‍♀️ Training LoRas

  • Read the following article for tips and my training preset

🌞 Final Remarks

  • This may be the last version of the model for now, as I am satisfied with the current results.
  • Thank you for downloading and using the model. I hope you enjoy it!

❤️ Support

  • If you use the site generator, you can support me increasing the Creator Tip.
  • Alternatively, you can support me on Ko-Fi.

⚡️ Buzz for the Best Images ⚡️

  • Improved lighting
  • Better performance in dark/night scenes
  • Slightly improved hands and feet
  • More refined colors
  • Cleaner overall image quality
  • Slightly better face shapes from different perspectives
  • Subtle improvements to skin
  • Enhanced outputs with Euler A
  • Improved prompt comprehension
  • Slightly finer details
  • Updated VAE

Content copied from Civitai: Pony Realism 🔮 on 2024-10-06 06:17:49.091744+00:00

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

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, load the workflow from the model's card on Hugging Face, and choose this file in its Load VAE node. 

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

pipe = DiffusionPipeline.from_pretrained("TheImposterImposters/PonyRealism-v2.2MainVAE", 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 (6.6 GB)
get TheImposterImposters/PonyRealism-v2.2MainVAE ponyRealism_v22MainVAE.safetensors vae

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