Spaces:
Running
on
Zero
Running
on
Zero
File size: 10,842 Bytes
d193c5b e1898cd 66fd925 d5f497d 66fd925 d193c5b e570df1 d193c5b e570df1 d193c5b e570df1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 |
---
title: FACE ID by Kolors Portrait
emoji: ๐ฅธ
colorFrom: purple
colorTo: yellow
sdk: gradio
sdk_version: 5.35.0
app_file: app.py
pinned: false
license: apache-2.0
---
## ๐จ Kolors IP-Adapter FaceID Plus: Advanced AI Portrait Generation
### Transform Faces into Stunning AI Artwork with Cutting-Edge Technology
Welcome to **Kolors IP-Adapter FaceID Plus**, a revolutionary AI portrait generation system that combines face recognition technology with advanced diffusion models. This powerful tool creates high-quality, personalized images by seamlessly blending facial features from reference photos with creative text prompts, opening new possibilities for digital art and portrait photography.
### What is Kolors IP-Adapter FaceID Plus?
Kolors IP-Adapter FaceID Plus is an **advanced AI image generation system** that leverages Kwai-Kolors' state-of-the-art models with InsightFace technology. It specializes in creating personalized portraits by extracting facial features from input images and incorporating them into AI-generated artwork based on text descriptions.
### Key Features for Professional Portrait Generation
- **๐ฏ Face Recognition Integration**: Advanced InsightFace technology for accurate facial feature extraction
- **๐ผ๏ธ High-Resolution Output**: Generate stunning 1024x1024 pixel portraits
- **๐จ Kolors Model**: Powered by Kwai-Kolors' advanced diffusion pipeline
- **๐ค Face Fidelity Control**: Adjustable face similarity strength (0.8 scale)
- **๐ง CLIP Vision Encoder**: Sophisticated facial feature understanding
- **โก GPU Acceleration**: Optimized for CUDA-enabled devices
- **๐ฒ Seed Control**: Reproducible results with manual or random seeds
- **๐ Negative Prompts**: Fine-tune output by excluding unwanted elements
### How It Works
#### **Intelligent Face Processing Pipeline**
1. **Face Detection**: AntelopeV2 model identifies and extracts facial features
2. **Face Embedding**: Converts facial characteristics into numerical representations
3. **Feature Integration**: Blends face embeddings with text prompts
4. **Image Generation**: Creates personalized portraits using diffusion technology
#### **Advanced Technical Components**
- **ChatGLM Text Encoder**: Sophisticated language understanding
- **UNet2D Conditional Model**: High-quality image generation
- **AutoencoderKL VAE**: Efficient latent space processing
- **Euler Discrete Scheduler**: Optimized sampling strategy
### Perfect Applications
- **Digital Portraits**: Create artistic interpretations of photographs
- **Character Design**: Generate consistent characters for stories or games
- **Avatar Creation**: Personalized profile pictures and digital identities
- **Fashion Visualization**: Try different styles and looks virtually
- **Historical Reimagining**: Place faces in different time periods or styles
- **Creative Photography**: Explore artistic portrait concepts
### Generation Parameters
- **Prompt**: Describe the desired scene, style, and atmosphere
- **Reference Image**: Upload a clear face photo for feature extraction
- **Negative Prompt**: Specify elements to avoid (default: "low quality")
- **Seed**: Control randomness (0-2147483647)
- **Guidance Scale**: Balance between prompt adherence and creativity (0-10)
- **Inference Steps**: Quality vs speed trade-off (10-50 steps)
### Technical Specifications
- **Base Model**: Kwai-Kolors/Kolors
- **Face Adapter**: Kolors-IP-Adapter-FaceID-Plus
- **Face Analysis**: InsightFace AntelopeV2
- **Image Encoder**: CLIP-ViT-Large-Patch14-336
- **Framework**: Diffusers, PyTorch, Gradio
- **Precision**: float16 for optimal performance
- **Resolution**: 1024x1024 output images
### Why Choose Kolors IP-Adapter FaceID Plus?
1. **Superior Face Preservation**: Maintains facial identity while allowing creative freedom
2. **Professional Quality**: Gallery-ready portraits with consistent results
3. **Easy to Use**: Simple interface with powerful underlying technology
4. **Versatile Styling**: From photorealistic to highly stylized outputs
5. **Fast Generation**: Optimized pipeline for quick results
6. **Open Technology**: Built on open-source foundations
### Best Practices for Portrait Generation
- **Clear Reference Photos**: Use well-lit, front-facing images for best results
- **Detailed Prompts**: Describe style, mood, lighting, and environment
- **Experiment with Seeds**: Try different seeds for varied interpretations
- **Adjust Guidance Scale**: Higher values follow prompts more strictly
- **Use Negative Prompts**: Exclude unwanted elements like "blurry, distorted"
- **Multiple Attempts**: Generate several versions to find the perfect result
### Privacy and Ethical Considerations
- Face data is processed locally during generation
- No permanent storage of facial information
- Use only with consent for any uploaded images
- Respect portrait rights and privacy laws
- Avoid creating misleading or harmful content
### Get Started with AI Portraits
Transform your photos into stunning AI artwork today. Whether creating personal portraits, character designs, or exploring artistic concepts, Kolors IP-Adapter FaceID Plus provides the tools for professional-quality results.
**Community**: [Discord - Openfree AI](https://discord.gg/openfreeai) | **More AI Services**: [OpenFree Best AI](https://huggingface.co/spaces/openfree/Best-AI)
---
## ๐จ Kolors IP-์ด๋ํฐ FaceID Plus: ๊ณ ๊ธ AI ์ด์ํ ์์ฑ
### ์ต์ฒจ๋จ ๊ธฐ์ ๋ก ์ผ๊ตด์ ๋๋ผ์ด AI ์์ ์ํ์ผ๋ก ๋ณํ
**Kolors IP-์ด๋ํฐ FaceID Plus**์ ์ค์ ๊ฒ์ ํ์ํฉ๋๋ค. ์ผ๊ตด ์ธ์ ๊ธฐ์ ๊ณผ ๊ณ ๊ธ ํ์ฐ ๋ชจ๋ธ์ ๊ฒฐํฉํ ํ์ ์ ์ธ AI ์ด์ํ ์์ฑ ์์คํ
์
๋๋ค. ์ด ๊ฐ๋ ฅํ ๋๊ตฌ๋ ์ฐธ์กฐ ์ฌ์ง์ ์ผ๊ตด ํน์ง์ ์ฐฝ์์ ์ธ ํ
์คํธ ํ๋กฌํํธ์ ์ํํ๊ฒ ํผํฉํ์ฌ ๊ณ ํ์ง์ ๊ฐ์ธํ๋ ์ด๋ฏธ์ง๋ฅผ ์์ฑํ๋ฉฐ, ๋์งํธ ์ํธ์ ์ด์ํ ์ฌ์ง์ ์๋ก์ด ๊ฐ๋ฅ์ฑ์ ์ด์ด์ค๋๋ค.
### Kolors IP-์ด๋ํฐ FaceID Plus๋?
Kolors IP-์ด๋ํฐ FaceID Plus๋ Kwai-Kolors์ ์ต์ฒจ๋จ ๋ชจ๋ธ๊ณผ InsightFace ๊ธฐ์ ์ ํ์ฉํ๋ **๊ณ ๊ธ AI ์ด๋ฏธ์ง ์์ฑ ์์คํ
**์
๋๋ค. ์
๋ ฅ ์ด๋ฏธ์ง์์ ์ผ๊ตด ํน์ง์ ์ถ์ถํ๊ณ ํ
์คํธ ์ค๋ช
์ ๊ธฐ๋ฐ์ผ๋ก AI ์์ฑ ์์ ์ํ์ ํตํฉํ์ฌ ๊ฐ์ธํ๋ ์ด์ํ๋ฅผ ๋ง๋๋ ๋ฐ ํนํ๋์ด ์์ต๋๋ค.
### ์ ๋ฌธ ์ด์ํ ์์ฑ์ ์ํ ์ฃผ์ ๊ธฐ๋ฅ
- **๐ฏ ์ผ๊ตด ์ธ์ ํตํฉ**: ์ ํํ ์ผ๊ตด ํน์ง ์ถ์ถ์ ์ํ ๊ณ ๊ธ InsightFace ๊ธฐ์
- **๐ผ๏ธ ๊ณ ํด์๋ ์ถ๋ ฅ**: ๋๋ผ์ด 1024x1024 ํฝ์
์ด์ํ ์์ฑ
- **๐จ Kolors ๋ชจ๋ธ**: Kwai-Kolors์ ๊ณ ๊ธ ํ์ฐ ํ์ดํ๋ผ์ธ ๊ตฌ๋
- **๐ค ์ผ๊ตด ์ถฉ์ค๋ ์ ์ด**: ์กฐ์ ๊ฐ๋ฅํ ์ผ๊ตด ์ ์ฌ์ฑ ๊ฐ๋ (0.8 ์ค์ผ์ผ)
- **๐ง CLIP ๋น์ ์ธ์ฝ๋**: ์ ๊ตํ ์ผ๊ตด ํน์ง ์ดํด
- **โก GPU ๊ฐ์**: CUDA ์ง์ ์ฅ์น์ ์ต์ ํ
- **๐ฒ ์๋ ์ ์ด**: ์๋ ๋๋ ๋ฌด์์ ์๋๋ก ์ฌํ ๊ฐ๋ฅํ ๊ฒฐ๊ณผ
- **๐ ๋ค๊ฑฐํฐ๋ธ ํ๋กฌํํธ**: ์ํ์ง ์๋ ์์๋ฅผ ์ ์ธํ์ฌ ์ถ๋ ฅ ๋ฏธ์ธ ์กฐ์
### ์๋ ๋ฐฉ์
#### **์ง๋ฅํ ์ผ๊ตด ์ฒ๋ฆฌ ํ์ดํ๋ผ์ธ**
1. **์ผ๊ตด ๊ฐ์ง**: AntelopeV2 ๋ชจ๋ธ์ด ์ผ๊ตด ํน์ง์ ์๋ณํ๊ณ ์ถ์ถ
2. **์ผ๊ตด ์๋ฒ ๋ฉ**: ์ผ๊ตด ํน์ฑ์ ์์น ํํ์ผ๋ก ๋ณํ
3. **ํน์ง ํตํฉ**: ์ผ๊ตด ์๋ฒ ๋ฉ์ ํ
์คํธ ํ๋กฌํํธ์ ํผํฉ
4. **์ด๋ฏธ์ง ์์ฑ**: ํ์ฐ ๊ธฐ์ ์ ์ฌ์ฉํ์ฌ ๊ฐ์ธํ๋ ์ด์ํ ์์ฑ
#### **๊ณ ๊ธ ๊ธฐ์ ๊ตฌ์ฑ ์์**
- **ChatGLM ํ
์คํธ ์ธ์ฝ๋**: ์ ๊ตํ ์ธ์ด ์ดํด
- **UNet2D ์กฐ๊ฑด๋ถ ๋ชจ๋ธ**: ๊ณ ํ์ง ์ด๋ฏธ์ง ์์ฑ
- **AutoencoderKL VAE**: ํจ์จ์ ์ธ ์ ์ฌ ๊ณต๊ฐ ์ฒ๋ฆฌ
- **Euler ์ด์ฐ ์ค์ผ์ค๋ฌ**: ์ต์ ํ๋ ์ํ๋ง ์ ๋ต
### ์๋ฒฝํ ์์ฉ ๋ถ์ผ
- **๋์งํธ ์ด์ํ**: ์ฌ์ง์ ์์ ์ ํด์ ์์ฑ
- **์บ๋ฆญํฐ ๋์์ธ**: ์ด์ผ๊ธฐ๋ ๊ฒ์์ ์ํ ์ผ๊ด๋ ์บ๋ฆญํฐ ์์ฑ
- **์๋ฐํ ์์ฑ**: ๊ฐ์ธํ๋ ํ๋กํ ์ฌ์ง ๋ฐ ๋์งํธ ์ ์
- **ํจ์
์๊ฐํ**: ๊ฐ์์ผ๋ก ๋ค์ํ ์คํ์ผ๊ณผ ๋ฃฉ ์๋
- **์ญ์ฌ์ ์ฌํด์**: ๋ค๋ฅธ ์๋๋ ์คํ์ผ์ ์ผ๊ตด ๋ฐฐ์น
- **์ฐฝ์์ ์ฌ์ง**: ์์ ์ ์ด์ํ ๊ฐ๋
ํ๊ตฌ
### ์์ฑ ๋งค๊ฐ๋ณ์
- **ํ๋กฌํํธ**: ์ํ๋ ์ฅ๋ฉด, ์คํ์ผ ๋ฐ ๋ถ์๊ธฐ ์ค๋ช
- **์ฐธ์กฐ ์ด๋ฏธ์ง**: ํน์ง ์ถ์ถ์ ์ํ ์ ๋ช
ํ ์ผ๊ตด ์ฌ์ง ์
๋ก๋
- **๋ค๊ฑฐํฐ๋ธ ํ๋กฌํํธ**: ํผํ ์์ ์ง์ (๊ธฐ๋ณธ๊ฐ: "low quality")
- **์๋**: ๋ฌด์์์ฑ ์ ์ด (0-2147483647)
- **๊ฐ์ด๋์ค ์ค์ผ์ผ**: ํ๋กฌํํธ ์ค์์ ์ฐฝ์์ฑ ๊ฐ์ ๊ท ํ (0-10)
- **์ถ๋ก ๋จ๊ณ**: ํ์ง ๋ ์๋ ๊ท ํ (10-50๋จ๊ณ)
### ๊ธฐ์ ์ฌ์
- **๊ธฐ๋ณธ ๋ชจ๋ธ**: Kwai-Kolors/Kolors
- **์ผ๊ตด ์ด๋ํฐ**: Kolors-IP-Adapter-FaceID-Plus
- **์ผ๊ตด ๋ถ์**: InsightFace AntelopeV2
- **์ด๋ฏธ์ง ์ธ์ฝ๋**: CLIP-ViT-Large-Patch14-336
- **ํ๋ ์์ํฌ**: Diffusers, PyTorch, Gradio
- **์ ๋ฐ๋**: ์ต์ ์ฑ๋ฅ์ ์ํ float16
- **ํด์๋**: 1024x1024 ์ถ๋ ฅ ์ด๋ฏธ์ง
### Kolors IP-์ด๋ํฐ FaceID Plus๋ฅผ ์ ํํด์ผ ํ๋ ์ด์
1. **์ฐ์ํ ์ผ๊ตด ๋ณด์กด**: ์ฐฝ์์ ์์ ๋ฅผ ํ์ฉํ๋ฉด์ ์ผ๊ตด ์ ์ฒด์ฑ ์ ์ง
2. **์ ๋ฌธ๊ฐ ํ์ง**: ์ผ๊ด๋ ๊ฒฐ๊ณผ๋ก ๊ฐค๋ฌ๋ฆฌ ์ค๋น ์ด์ํ
3. **์ฌ์ฉํ๊ธฐ ์ฌ์**: ๊ฐ๋ ฅํ ๊ธฐ๋ณธ ๊ธฐ์ ์ ๊ฐ์ถ ๊ฐ๋จํ ์ธํฐํ์ด์ค
4. **๋ค์ํ ์คํ์ผ๋ง**: ์ฌ์ค์ ์ธ ๊ฒ๋ถํฐ ๊ณ ๋๋ก ์์ํ๋ ์ถ๋ ฅ๊น์ง
5. **๋น ๋ฅธ ์์ฑ**: ๋น ๋ฅธ ๊ฒฐ๊ณผ๋ฅผ ์ํ ์ต์ ํ๋ ํ์ดํ๋ผ์ธ
6. **๊ฐ๋ฐฉํ ๊ธฐ์ **: ์คํ์์ค ๊ธฐ๋ฐ์ผ๋ก ๊ตฌ์ถ
### ์ด์ํ ์์ฑ ๋ชจ๋ฒ ์ฌ๋ก
- **์ ๋ช
ํ ์ฐธ์กฐ ์ฌ์ง**: ์ต์์ ๊ฒฐ๊ณผ๋ฅผ ์ํด ์กฐ๋ช
์ด ์ข๊ณ ์ ๋ฉด์ ํฅํ ์ด๋ฏธ์ง ์ฌ์ฉ
- **์์ธํ ํ๋กฌํํธ**: ์คํ์ผ, ๋ถ์๊ธฐ, ์กฐ๋ช
๋ฐ ํ๊ฒฝ ์ค๋ช
- **์๋ ์คํ**: ๋ค์ํ ํด์์ ์ํด ๋ค๋ฅธ ์๋ ์๋
- **๊ฐ์ด๋์ค ์ค์ผ์ผ ์กฐ์ **: ๋์ ๊ฐ์ ํ๋กฌํํธ๋ฅผ ๋ ์๊ฒฉํ๊ฒ ๋ฐ๋ฆ
- **๋ค๊ฑฐํฐ๋ธ ํ๋กฌํํธ ์ฌ์ฉ**: "blurry, distorted"์ ๊ฐ์ ์ํ์ง ์๋ ์์ ์ ์ธ
- **์ฌ๋ฌ ์๋**: ์๋ฒฝํ ๊ฒฐ๊ณผ๋ฅผ ์ฐพ๊ธฐ ์ํด ์ฌ๋ฌ ๋ฒ์ ์์ฑ
### ๊ฐ์ธ์ ๋ณด ๋ณดํธ ๋ฐ ์ค๋ฆฌ์ ๊ณ ๋ ค์ฌํญ
- ์ผ๊ตด ๋ฐ์ดํฐ๋ ์์ฑ ์ค์ ๋ก์ปฌ๋ก ์ฒ๋ฆฌ๋จ
- ์ผ๊ตด ์ ๋ณด์ ์๊ตฌ ์ ์ฅ ์์
- ์
๋ก๋๋ ์ด๋ฏธ์ง์ ๋ํด์๋ง ๋์ํ์ ์ฌ์ฉ
- ์ด์๊ถ ๋ฐ ๊ฐ์ธ์ ๋ณด ๋ณดํธ๋ฒ ์ค์
- ์คํด์ ์์ง๊ฐ ์๊ฑฐ๋ ํด๋ก์ด ์ฝํ
์ธ ์์ฑ ๊ธ์ง
### AI ์ด์ํ ์์ํ๊ธฐ
์ค๋ ์ฌ์ง์ ๋๋ผ์ด AI ์์ ์ํ์ผ๋ก ๋ณํํ์ธ์. ๊ฐ์ธ ์ด์ํ, ์บ๋ฆญํฐ ๋์์ธ ๋๋ ์์ ์ ๊ฐ๋
ํ๊ตฌ ๋ฑ ๋ฌด์์ ๋ง๋ค๋ Kolors IP-์ด๋ํฐ FaceID Plus๋ ์ ๋ฌธ๊ฐ ์์ค์ ๊ฒฐ๊ณผ๋ฅผ ์ํ ๋๊ตฌ๋ฅผ ์ ๊ณตํฉ๋๋ค.
**์ปค๋ฎค๋ํฐ**: [Discord - Openfree AI](https://discord.gg/openfreeai) | **๋ ๋ง์ AI ์๋น์ค**: [OpenFree Best AI](https://huggingface.co/spaces/openfree/Best-AI) |