Rev Engine PonyXL

세부 정보

모델 설명

This merge is inspired by Rev Animated - one of the favorite stylized checkpoints for SD 1.5.

Built on top of DucHaiten, Everclear and several carefully selected Loras to achieve that distinctive artistic style.

Base Model: SDXL 1.0 / Pony Diffusion V6 XL

Type: Stylized Checkpoint Merge

✨Specific:

  • Works great for artistic and stylized generations

  • Produces very unique aesthetic results when using LoRAs

  • VAE installation is mandatory for proper output


⚠️ I M P O R T A N T !

Requires SDXL VAE, otherwise images will be washed out with white dots.

  • Recommended VAE:

Pony Standard VAE - specifically designed for Pony/SDXL models

https://civitai.com/models/660673/pony-standard-vae


📝 Note:

This is a reprint of a previously available model. Original upload by razzz (Jul, 2024) was removed by the creator, but the model remains popular in the community.


⚙️ Recommended Generation Settings:

🏆 Main Settings:

  • Sampling steps: 30

  • Sampling method: DPM++ 2M SDE

  • Schedule type: Karras

  • CFG Scale: 2-4

🏷️ Hires.fix Settings:

  • Upscaler: R-ESRGAN 4x+ Anime6B

  • Hires steps: 20

  • Denoising strength: 0.35

  • Hires Distilled CFG Scale: 3.5

  • Hires CFG Scale: 1.5-2


📋 Recommended Prompt Structure:

[ Mandatory Quality Tags ]

  • score_9, score_8_up, score_7_up, score_6_up,

  • source_photo, realistic, 1girl, full body

[ DanBooru Tags ]

  • Use DanBooru-style tags here

  • For img2img: Use WD14 Tagger with model wd14-vit-v2-git

https://github.com/Akegarasu/sd-webui-wd14-tagger

[ CLIP Description (optional) ]

  • Natural language description

  • For img2img: Use CLIP Interrogator with model ViT-L-14/openai

https://github.com/pharmapsychotic/clip-interrogator-ext

[ LORA ]

  • Add your desired LoRAs here


🎨 Recommended LoRAs:

💡 Tip: Combine 3-4 LoRAs from different groups.

Keep total weight ≤ 1.5-2.0 for best results.

🎭 Group 1 - Style:

💡 Group 2 - Lighting:

🖼️ Group 3 - Concept:

😊 Group 4 - Expression:

🌈 Group 5 - Color/Theme:


🚫 Recommended Negative Embedding:


❤️ Artificial Inspiration by DORO

이 모델로 만든 이미지