---
title: Imagen
emoji: 🌖
colorFrom: pink
colorTo: red
sdk: gradio
sdk_version: 5.22.0
app_file: main.py
pinned: false
license: other
short_description: Generate and edit images
---

# Diffusion Models App

A unified interface for text-to-image and image-to-image generation using Hugging Face models.

## Features

- 🖼️ **Text to Image**: Generate images from text prompts
- 🔄 **Image to Image**: Transform images using text prompts
- 🧠 **ControlNet Support**: Built-in ControlNet depth model for enhanced image transformations
- 🌐 **Flexible Models**: Use inference endpoints or on-device models
- 🚀 **Dual Interface**: Web UI and API endpoints

## Deployment Notes

### IMPORTANT: Using ControlNet on Hugging Face Spaces

For ControlNet to work correctly:

1. Import `spaces` before any `torch` or CUDA-related imports
2. Select GPU hardware in Space settings
3. Add your HF_TOKEN as a Repository Secret (Settings → Repository Secrets)
4. Do NOT commit any sensitive tokens to the repository

If you encounter CUDA initialization errors, ensure spaces package is imported first.

## Usage

### Text to Image
1. Enter your text prompt
2. Optionally provide a negative prompt to exclude unwanted elements
3. Choose a model or use the default
4. Click "Generate Image"

### Image to Image
1. Upload an image
2. Enter a prompt to guide the transformation
3. Choose between HF inference API or ControlNet on-device model
4. Click "Transform Image"

---

tags:
- diffusers
- stable-diffusion
- text-to-image
- image-to-image
- depth-estimation
- controlnet
- spaces
- huggingface-hub