Hugging Face's logo

Spaces:
lulavc
/
Z-Image-Turbo
Running on Zero

App Files
Z-Image-Turbo / README.md
lulavc's picture
lulavc
Upload README.md with huggingface_hub
7238837 verified
preview code
|
raw
history blame
2.78 kB
metadata
title: Z-Image Turbo + GLM-4.6V / DeepSeek-3.2 Thinking
emoji: 🎨
colorFrom: purple
colorTo: pink
sdk: gradio
sdk_version: 5.31.0
app_file: app.py
pinned: true
license: apache-2.0
tags:
  - text-to-image
  - image-to-image
  - diffusion
  - transformer
  - z-image
  - z-image-turbo
  - ai-art
  - image-generation
  - image-editing
  - glm-4
  - glm-4.6v
  - glm-4v
  - vision-language
  - multimodal
  - prompt-generation
  - ai-assistant
  - zhipu
  - zai
  - deepseek
  - deepseek-reasoner
  - deepseek-3.2
  - thinking
  - reasoning
  - multilingual
  - rtl
  - arabic
  - hindi
  - spanish
  - portuguese
short_description: Image Gen & Transform. Van Gogh and Picasso style included!

⚡ Z-Image Generation & Transformation Demo

Generate stunning AI images and transform existing photos using Alibaba's Z-Image-Turbo - a next-generation diffusion transformer model.

✨ Features

🎨 Text-to-Image Generation

  • 12 style presets (Photorealistic, Cinematic, Anime, Van Gogh, Picasso, and more)
  • 18 aspect ratios from 1024px to 2048px
  • AI-powered prompt polishing with DeepSeek Reasoner

🖼️ Image-to-Image Transformation

  • Transform photos into different art styles
  • Van Gogh style: Post-impressionist swirling brushstrokes, thick impasto, vibrant colors
  • Picasso style: Cubist geometric fragmentation, multiple perspectives, bold outlines
  • Style-only transformation: just select a style, no prompt needed!
  • Adjustable strength control (subtle to complete change)

⚡ Optimized Performance

  • Runs on ZeroGPU (NVIDIA H200)
  • FlashAttention-3 via kernels library
  • TF32 + VAE torch.compile optimization
  • ~20-25 seconds per generation

🚀 How to Use

Generate Tab

  1. Enter a description of your image
  2. (Optional) Enable "Prompt+" for AI-enhanced prompts
  3. Select a style and aspect ratio
  4. Click "Generate"

Transform Tab

  1. Upload an image
  2. Select a style (Van Gogh, Picasso, etc.) - no prompt needed!
  3. Or add a custom prompt for more control
  4. Adjust strength (lower = subtle, higher = dramatic)
  5. Click "Transform"

🎯 Example Prompts

  • "Ancient dragon perched on a Gothic cathedral at dusk, stormy purple sky"
  • "Cyberpunk samurai in a neon-lit rainy alley, glowing armor"
  • "Cozy witch cottage interior, cauldrons bubbling, magical atmosphere"

📊 Technical Details

Spec Value
Model Tongyi-MAI/Z-Image-Turbo
Steps 9 (optimized)
Precision BFloat16
Max Resolution 2048x2048

🔗 Links

📜 License

Model: Apache 2.0 License

Created by @lulavc