Anything2Real

提供商lrzjason
分类image-to-image
许可证apache-2.0
下载量161.0K
星标0

简介

Anything2Real 是一款专注于将各类图像转化为写实风格的图生图模型。它解决了许多 AI 绘画中常见的“塑料感”或过度动漫化问题，能将简单的草图、动漫角色或合成图像转化为具有真实摄影质感的照片。对于习惯使用 Stable Diffusion 的开发者和创作者来说，它像是一个强大的写实滤镜插件，极大地降低了通过复杂 Prompt 追求真实感的难度，非常适合需要快速出写实效果图的电商、原画设计或社交媒体创作场景。

核心亮点

  • 高效将动漫/草图转化为高保真写实照片
  • 显著降低写实风格 Prompt 的调试成本
  • 适配 Stable Diffusion 生态，上手门槛低
  • Apache-2.0 协议，支持商业化灵活部署

使用方法

安装依赖
# 安装 Hugging Face transformers
pip install transformers torch
SDK 使用
# 使用 transformers 加载模型
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained("lrzjason/Anything2Real")
tokenizer = AutoTokenizer.from_pretrained("lrzjason/Anything2Real")

Hugging Face 下载

我们推荐使用命令行或者 Hugging Face Hub SDK 来进行模型的下载。

操作指引：在下载前，请先通过如下命令安装 huggingface_hub：

操作指引 
pip install -U huggingface_hub

命令行下载

下载完整模型库

下载完整模型库 
huggingface-cli download lrzjason/Anything2Real

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例）

下载单个文件到指定本地文件夹（以下载 config.json 到当前路径下 ./dir 目录为例） 
huggingface-cli download lrzjason/Anything2Real config.json --local-dir ./dir

更多命令行下载选项，可参见官方文档

SDK 下载

SDK 下载 
# 模型下载
from huggingface_hub import snapshot_download
model_dir = snapshot_download('lrzjason/Anything2Real')

Git 下载

请确保 lfs 已经被正确安装

Git 下载 
git lfs install
git clone https://huggingface.co/lrzjason/Anything2Real

如果您希望跳过 lfs 大文件下载，可以使用如下命令

跳过 LFS 
GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/lrzjason/Anything2Real

模型文件托管在 Hugging Face Hub，使用 HF CLI / SDK / Git 直接下载，不经过本站。

PyTorch / Transformers 使用

安装 Transformers

安装 Transformers 
pip install -U transformers torch

模型加载和推理

模型加载和推理 
from transformers import AutoModelForCausalLM, AutoTokenizer

model = AutoModelForCausalLM.from_pretrained('lrzjason/Anything2Real')
tokenizer = AutoTokenizer.from_pretrained('lrzjason/Anything2Real')

完整文档

来源: HuggingFace

---
license: apache-2.0
base_model:

  • Qwen/Qwen-Image-Edit-2511

library_name: diffusers
tags:
  • lora

pipeline_tag: image-to-image
---

[RELEASE] New Version of Anything2Real LoRA - Transform Any Art Style to Photorealistic Images Based On Flux2 Klein

f2k_anything2real.safetensors is more 3D like version
f2k_anything2real_a.safetensors based on f2k_anything2real.safetensors to adjust the realistic style

<table>
<tr>
<td>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/QTBiWwlzKul23Mo-kG-iY.png" style="max-width:300px"/>
</td>
<td>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/SCJwYwmBYDM_itlCzQ7x8.png" style="max-width:300px"/>
</td>
</tr>
</table>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/plOqXDhtipDimuk5LMwha.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/AzhZjNSXvO2yhPDfG6BTs.png" style="max-width:600px"/>

[RELEASE] New Version of Anything2Real LoRA - Transform Any Art Style to Photorealistic Images Based On Qwen Edit 2511

<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/YGQuZHBUsxJ1BIM_iMGbu.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/TCN_KuvmGrOJ_eKevCc0G.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/29VMNd6KWNoJYbCSYr_V5.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/Y1MOrhM9p7MRRsqEmi2rw.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/AOFw8QXJHECN22iJeuRvX.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/leeP3Vbr4fMtF7JhmC0E6.png" style="max-width:600px"/>
<img src="https://cdn-uploads.huggingface.co/production/uploads/63891deed68e37abd59e883f/3ylW_7t3U6yEoWeFnK8Rz.png" style="max-width:600px"/>

Hey Stable Diffusion community! 👋

I'm excited to share the new version of Anything2Real family, Anything2Real 2601 A.
Anything2Real 2601 A is an adjusted version based on Anything2Real 2601 and remixed the layer weight of anything2real alpha.
Anything2Real 2601 is kinda weak of eyes transformation and anything2real alpha is quite good at light and shadow but weak at color.
Anything2Real 2601 A is a more balanced version between "respect to original" and "realistic environemnt"

Alpha Dataset: lrzjason/Anything2Real_demo_dataset

🔗 Civitai: https://civitai.com/models/2121900/qwen-edit-2511-anything2real-2601

🎯 What It Does

This LoRA is designed to convert illustrations, anime, cartoons, paintings, and other non-photorealistic images into convincing photographs while preserving the original composition and content.

⚙️ How to Use

  • Base Model: Qwen Edit 2511 (mmdit editing model)
  • Recommended Strength: 1-0.75
  • Prompt Template:
code
change the picture 1 to realistic photograph, [description of your image]
  • Adding detailed descriptions helps the model better understand content and produces superior transformations (though it works even without detailed prompts!)

📺 Example Workflow

Check out this practical implementation example on RunningHub: https://www.runninghub.ai/post/1988649905554112513/?inviteCode=rh-v1279

Node Information:

  • 21 nodes including CLIPLoader, ConditioningZeroOut, ImageScaleBy, KSampler, LoadImage, LoraLoaderModelOnly, and specialized Qwen Edit nodes

  • Powered by RTX 4090 Series

📌 Important Notes

  • “realism” is inherently subjective, first modulate strength or switch base models rather than further increasing the LoRA weight.
  • Should realism remain insufficient, blend with an additional photorealistic LoRA and adjust to taste.
  • Your feedback and examples would be incredibly valuable for future improvements!

💡 Why I Created This

I wanted to bridge the gap between artistic creations and photorealistic results without having to start from scratch. Whether you're an artist wanting to see your characters in realistic form, or just curious about transforming your favorite art styles into photographs, this tool is for you!

I'd love to see what you create with Anything2Real! Please share your results and suggestions in the comments. Every test case helps improve the next version.

---

*Note: For best results, follow the workflow demonstrated in the RunningHub example. Feel free to reach out with questions or collaboration ideas!*

Contact

Feel free to reach out via any of the following channels: Twitter: @Lrzjason Email: [email protected] QQ Group: 866612947 WeChat ID: fkdeai CivitAI: xiaozhijason
查看官方来源