AI Image Generation Platform
🎨 Stable Diffusion | 🌍 Web Interface | ⚡ Fast | 🎯 Precise | 🖼️ Gallery | 🔄 Queue System | 📡 API | 🤖 MCP Support
| 🎨 Image Generation | 🖥️ Web Interface | 🖼️ Gallery Features | ⚡ Performance |
|---|---|---|---|
| 🤖 State-of-the-art AI models | 🌃 Cyberpunk-themed UI | 📊 Multiple view options | 💾 Smart model caching |
| 🎛️ Customizable parameters | ⏱️ Real-time status updates | 🔍 Detailed image modal | 🧠 Memory optimization |
| 🚫 Negative prompt support | 🖱️ Interactive gallery | ⬇️ Quick download actions | 🔄 Efficient queue system |
| 📦 Batch image generation | 📱 Mobile-responsive design | 📋 Copy prompt feature | 🏃♂️ Background processing |
| 📈 Real-time progress | 🌈 Beautiful UI | 🔍 Search and filtering | 🔒 No data collection |
| 🎯 Precise control | 🎮 Intuitive controls | 🏷️ Tagging system | 🏠 Local image storage |
| 🧩 Model compatibility | 🌙 Dark mode support | ||
| 🤖 MCP Integration | 🔌 AI Accessibility | ||
| 🔗 AI assistant support | 🔄 JSON-RPC interface |
CyberImage now implements the Model Context Protocol (MCP), enabling AI assistants and other tools to seamlessly generate images through a standardized interface.
MCP is an open protocol that enables AI systems to interact with external tools and capabilities in a standardized way. With MCP support, AI assistants can generate images directly through CyberImage using JSON-RPC calls.
- JSON-RPC 2.0 Interface: Simple, standardized format for all requests
- Dynamic Default Model: Uses the system's default model (same as the web UI), prioritizing "flux-1" if available
- Seamless Queue Integration: Jobs from AI assistants are integrated into the same queue as web UI requests
- Progress Tracking: AI systems can track generation progress in real-time
- Standard Format: Follows the MCP specification for interoperability with any MCP-compatible AI system
| Method | Description |
|---|---|
context.image_generation.models |
List all available models |
context.image_generation.generate |
Generate images based on a prompt |
context.image_generation.status |
Check the status of a generation job |
AI assistants can connect to the MCP endpoint at:
http://localhost:5050/api/mcp
For implementation examples, see the examples/ directory:
mcp_client_example.py: General MCP client implementationai_assistant_mcp_example.py: Specialized client for AI assistants
The Enrich button uses the openai api to enhance the image. It uses the openai api key and the openai model to enhance the image. The openai model is the same as the model used for generation.
This is my favorite feature, it allows you to take a basic image prompt and enhance it to make it better using a number of different techniques shared with myself by an AI expert.
| Requirement | Specification |
|---|---|
| Python | 3.12 (for local installation) |
| GPU | CUDA-capable |
| RAM | 16GB+ recommended |
| VRAM | 24GB+ recommended |
| Disk Space | 250GB+ for models |
| Container | Docker & Docker Compose (for containerized installation) |
| API | Huggingface API Key (free! for downloading models) |
- Copy the example environment file:
cp .env.example .env- Configure the environment variables in
.env:
-
For a local install, you can use the MODEL_FOLDER and IMAGES_FOLDER to store the models and images locally in different directories.
-
For a docker install, you can use the EXTERNAL_MODEL_FOLDER and EXTERNAL_IMAGES_FOLDER to store the models and images externally in different directories.
-
For a docker install, you will need a Huggingface API Key to download the models.
-
The openai endpoints works fine with Ollama with 127.0.0.1:11434/v1 as the endpoint and the openai api key as the key, or host.docker.internal:11434/v1 as the endpoint and the openai api key as the key if ollama is running on the host machine. The model needs to be something in 127.0.0.1:11434/v1/models on your system.
If you don't configure the openai endpoint, the enrich prompt will not work.
- The civitai api key is optional, it is used to download models from civitai (not configured currently).
MODEL_FOLDER=./models
IMAGES_FOLDER=./images
EXTERNAL_MODEL_FOLDER=
EXTERNAL_IMAGES_FOLDER=
HF_TOKEN=
OPENAI_ENDPOINT=
OPENAI_API_KEY=
OPENAI_MODEL=
CIVITAI_API_KEY=- Clone the repository:
git clone https://github.com/ramborogers/cyberimage.git
cd cyberimage- Use the run.sh script to start the application (easiest):
# This will start the application in a container
./run.sh startAlternatively, use Docker or Docker Compose directly:
- Using docker-compose:
# Start the application
docker-compose up -d
# View logs
docker-compose logs -f
# Stop the application
docker-compose down- Using Docker CLI:
# Build the image
docker build -t cyberimage .
# Run the container
docker run -d \
--name cyberimage \
--gpus all \
-p 7860:5050 \
-v $(pwd)/models:/app/models \
-v $(pwd)/images:/app/images \
--env-file .env \
cyberimage- Open in browser:
http://localhost:7860
- Clone the repository:
git clone https://github.com/ramborogers/cyberimage.git
cd cyberimage- Create a virtual environment:
python -m venv venv
source venv/bin/activate # Linux/Mac
# or
.\venv\Scripts\activate # Windows- Install dependencies:
pip install -r requirements.txt- Download models:
python download_models.py- Model: Choose from multiple AI models
- Prompt: Describe your desired image
- Negative Prompt: Specify elements to exclude
- Size: Select output dimensions
- Steps: Control generation quality
- Guidance: Adjust prompt adherence
- Batch Size: Generate multiple images
- AI Assistant Integration: Allow AI assistants to generate images based on user conversations
- Family Images: My children love to use this
- Digital Art: Create unique artwork
- Concept Design: Generate design ideas
- Visual Inspiration: Explore creative concepts
- Content Creation: Generate visual content
Copy the .env.example file to .env and edit the .env file to configure the models you want to use.
CyberImage uses environment variables in the .env file to configure models. You can easily add, remove, or modify models by editing this file.
Models are defined using the following format:
MODEL_<N>=<name>;<repo_or_url>;<description>;<source>;<requires_auth>[;<options_json>]
Where:
<N>: Numerical index (1, 2, 3, etc.). Must be unique.<name>: Unique identifier for the model (used as directory name).<repo_or_url>: HuggingFace repository path, GGUF file URL, or other model identifier.<description>: Human-readable description shown in the UI.<source>: Source platform (huggingface,gguf_url,civitai, etc.). Used bydownload_models.py.<requires_auth>: Whether authentication (e.g., HuggingFace token) is required (true/false).<options_json>(Optional): A JSON string containing additional model-specific parameters.{"fixed_steps": N}: Forces the model to use a specific number of steps (e.g., forsana-sprint).{"type": "t2v"}: Explicitly marks the model as Text-to-Video.{"type": "i2v"}: Explicitly marks the model as Image-to-Video.- Note: The application also attempts auto-detection based on model names (e.g., "t2v", "i2v"). Explicit types take precedence.
- To add: Add a new
MODEL_<N>line with a unique index and setDOWNLOAD_MODEL_<N>=true. - To disable download: Set
DOWNLOAD_MODEL_<N>=false. The model remains in the UI if already downloaded. - To remove: Delete or comment out both the
MODEL_<N>andDOWNLOAD_MODEL_<N>lines.
Here's an example .env configuration showing various model types:
# --- Image Models ---
MODEL_1="flux-1;black-forest-labs/FLUX.1-dev;FLUX Dev;huggingface;true"
DOWNLOAD_MODEL_1=true
MODEL_2="sd-3.5;stabilityai/stable-diffusion-3.5-large;Stable Diffusion 3.5;huggingface;true"
DOWNLOAD_MODEL_2=true
MODEL_3="flux-schnell;black-forest-labs/FLUX.1-schnell;FLUX Schnell;huggingface;true"
DOWNLOAD_MODEL_3=true
MODEL_4="sana-sprint;Efficient-Large-Model/Sana_Sprint_1.6B_1024px_diffusers;Sana Sprint 1.6B (Fixed Steps);huggingface;false;{\\"fixed_steps\\": 2}"
DOWNLOAD_MODEL_4=true
# --- Video Models ---
# Text-to-Video (Wan)
MODEL_5="wan-t2v-1.3b;Wan-AI/Wan2.1-T2V-1.3B-Diffusers;Wan Text2Video 1.3B;huggingface;false;{\\"type\\": \\"t2v\\"}"
DOWNLOAD_MODEL_5=true
# Image-to-Video (Wan)
MODEL_6="wan-i2v-14b;Wan-AI/Wan2.1-I2V-14B-480P;Wan Image-to-Video (14B, 480p);huggingface;false;{\\"type\\": \\"i2v\\"}"
DOWNLOAD_MODEL_6=true
# Text-to-Video (LTX GGUF - requires download_models.py support for gguf_url)
MODEL_7='LTX-Video-t2v;https://huggingface.co/city96/LTX-Video-gguf/resolve/main/ltx-video-2b-v0.9-Q3_K_S.gguf;LTX Video GGUF T2V;gguf_url;false;{"type": "t2v"}'
DOWNLOAD_MODEL_7=true
# Image-to-Video (LTX GGUF - uses same file as above)
MODEL_8='LTX-Video-i2v;https://huggingface.co/city96/LTX-Video-gguf/resolve/main/ltx-video-2b-v0.9-Q3_K_S.gguf;LTX Video GGUF I2V;gguf_url;false;{"type": "i2v"}'
DOWNLOAD_MODEL_8=false # Set to false if MODEL_7 already downloads the file
# --- Other Examples ---
# Disabled Model
MODEL_9="animagine-xl;cagliostrolab/animagine-xl-4.0;Animagine XL;huggingface;true"
DOWNLOAD_MODEL_9=false
# Commented out/Removed Model (example for Civitai - requires API key and download_models.py support)
# MODEL_10="my-custom-model;civitai:12345;My Custom Model;civitai;true"
# DOWNLOAD_MODEL_10=true(Remember to restart the application after changing .env)
CyberImage extends beyond static images, offering powerful Text-to-Video (T2V) and Image-to-Video (I2V) generation capabilities!
- Text-to-Video (T2V): Generate videos directly from text prompts using the main generation interface.
- Image-to-Video (I2V): Bring your existing generated images to life by creating videos from them, initiated directly from the gallery.
- Supported Models: Leverages advanced video models like
WanPipeline,WanImageToVideoPipeline, and experimentally supportsLTXPipeline(including GGUF variants). - Seamless Integration: Video jobs are handled by the same robust queue system as image jobs.
- Configuration: Add and manage video models via the
.envfile as described in the Managing Models section above. The application identifies video models using the optionaltypeparameter in the model configuration or by keywords in the model name (e.g., "t2v", "i2v").
- Text-to-Video:
- Select a T2V model (identified by name or
[Video]tag) in the main model dropdown. - Enter your prompt.
- Click "🎬 Generate Video".
- Select a T2V model (identified by name or
- Image-to-Video:
- Go to the Gallery or view a single image.
- Click the 🎥 (Generate Video) icon on the desired image.
- A modal will appear; enter a video prompt and select an available I2V model.
- Submit the job.
Generated videos will appear in the gallery alongside images, with appropriate video player controls.
- Fork the repository
- Create a feature branch
- Commit your changes
- Push to the branch
- Create a Pull Request
CyberImage is licensed under the GNU General Public License v3.0 (GPLv3).
Free Software
Made with 💚 by [Matthew Rogers]