atelier-generator 25.3.1

A comprehensive toolkit for state-of-the-art AI image generation compatible with all devices.

atelier-generator

A comprehensive toolkit for state-of-the-art AI image generation compatible with all devices.

Installation

pip install atelier-generator

Key Features

• 🎨 Image Generation
  • Text-to-Image Generation
  • Image Variations
  • Structural & Facial Guidance
  • Style Transfer & ControlNet
• 🛠️ Image Editing
  • Face Enhancement (GFPGAN/CodeFormer)
  • Background Removal
  • Image Upscaling
  • Object Erasing & Inpainting
• ⚡ Real-time Features
  • RT Image Generation
  • Interactive Canvas
  • Image Outpainting
  • Image Analysis (Caption/Prompt)

Usage

Python Library

from atelier_generator import AtelierGenerator

# Initialize
atelier = AtelierGenerator(
    mode="default", # Mode (default/webui/api)
    gradio=False, # Enable Gradio support
    timeout=180, # Request timeout (seconds)
    log_on=True, # Enable logging
    save_to="outputs", # Output directory
    save_as="webp" # Output format (png/webp/jpg/pil)
)

# Basic image generation
result = atelier.image_generate(
    prompt="a beautiful landscape",
    negative_prompt="", # Optional negative prompt
    model_name="flux-turbo", # Model selection
    image_size="1:1", # Output size ratio
    lora_svi=None, # LoRA SVI preset
    lora_flux=None, # LoRA Flux preset
    image_seed=0, # Generation seed
    style_name=None # Style preset
)

# Image variation
result = atelier.image_variation(
    image="source.jpg", # Source image
    prompt="convert to anime",
    negative_prompt="", # Optional negative prompt
    model_name="flux-turbo", # Model selection
    image_size="1:1", # Output size ratio
    strength="high", # Variation strength (low/medium/high)
    lora_svi=None, # LoRA SVI preset
    lora_flux=None, # LoRA Flux preset
    image_seed=0, # Generation seed
    style_name=None # Style preset
)

# Structural guidance
result = atelier.image_structure(
    image="structure.jpg", # Source image
    prompt="enhance details",
    negative_prompt="", # Optional negative prompt
    model_name="svi-realistic", # Model selection
    image_size="1:1", # Output size ratio
    strength="high", # Guide strength (low/medium/high)
    lora_svi=None,  # LoRA SVI preset
    image_seed=0, # Generation seed
    style_name=None # Style preset
)

# Face enhancement
result = atelier.face_gfpgan(
    image="face.jpg",
    model_version="1.3" # Model version (1.2/1.3)
)
result = atelier.face_codeformer(
    image="face.jpg"
)

# Image editing
result = atelier.image_enhance(
    image="photo.jpg",
    prompt="enhance quality", # Optional prompt
    negative_prompt="", # Optional negative prompt
    creativity=0.3, # Creativity level (0.0-1.0)
    resemblance=1.0, # Resemblance level (0.0-1.0)
    hdr=0.0, # HDR strength (0.0-1.0)
    style_name=None # Style preset
)

result = atelier.image_inpaint(
    image="image.jpg",
    mask="mask.jpg", # Mask image
    prompt="fill with trees",
    style_name=None # Style preset
)

result = atelier.image_erase(
    image="image.jpg",
    mask="mask.jpg" # Mask image
)

result = atelier.image_bgremove(
    image="photo.jpg"
)

result = atelier.image_upscale(
    image="small.jpg"
)

result = atelier.image_outpaint(
    image="image.jpg",
    image_size="16:9" # Output size ratio
)

# Real-time features
result = atelier.realtime_generate(
    prompt="quick sketch",
    negative_prompt="", # Optional negative prompt
    image_size="1:1", # Output size ratio
    lora_rt=None, # LoRA RT preset
    image_seed=0, # Generation seed
    style_name=None # Style preset
)

result = atelier.realtime_canvas(
    image="canvas.jpg", # Source image
    prompt="enhance drawing",
    negative_prompt="", # Optional negative prompt
    lora_rt=None, # LoRA RT preset
    strength=0.9, # Creativity level (0.0-1.0)
    image_seed=0, # Generation seed
    style_name=None # Style preset
)

# Image analysis
caption = atelier.image_caption(
    image="photo.jpg"
)
prompt = atelier.image_prompt(
    image="photo.jpg"
)

# ControlNet features
result = atelier.image_controlnet(
    image="sketch.jpg",
    prompt="convert to art",
    negative_prompt="", # Optional negative prompt
    model_name="sd-toon", # Model selection
    controlnet="scribble", # Control type (scribble/pose/line-art/depth/canny)
    strength=70, # Control strength (0-100)
    cfg=9.0, # Prompt guidance scale
    image_seed=0, # Generation seed
    style_name=None # Style preset
)

Web UI

Start the Gradio web interface:

atelier = AtelierGenerator(mode="webui")
# OR
atelier = AtelierGenerator()
atelier.start_wui(
    host="0.0.0.0", # Server host
    port=5735, # Server port
    browser=True, # Launch browser
    upload_size="4MB", # Max upload size
    public=False, # Enable public URL
    limit=10 # Max concurrent requests
)

REST API

Start the Flask API server:

atelier = AtelierGenerator(mode="api")
# OR
atelier = AtelierGenerator()
atelier.start_api(
    host="0.0.0.0", # Server host
    port=5733, # Server port
    debug=False # Enable debug mode
)

API Endpoints

Image Generation:

• POST /v1/api/image/generate - Generate images from text
• POST /v1/api/image/variation - Create image variations
• POST /v1/api/image/structure - Apply structural guidance
• POST /v1/api/image/facial - Apply facial guidance
• POST /v1/api/image/style - Apply style transfer
• POST /v1/api/image/controlnet - Apply ControlNet

Image Editing:

• POST /v1/api/image/enhance - Enhance image quality
• POST /v1/api/image/inpaint - Fill masked areas
• POST /v1/api/image/erase - Remove objects
• POST /v1/api/image/upscale - Upscale image
• POST /v1/api/image/bgremove - Remove background
• POST /v1/api/image/outpaint - Extend image borders

Face Enhancement:

• POST /v1/api/face/gfpgan - GFPGAN face restoration
• POST /v1/api/face/codeformer - CodeFormer face restoration

Real-time Features:

• POST /v1/api/realtime/generate - Real-time generation
• POST /v1/api/realtime/canvas - Interactive canvas

Image Analysis:

• POST /v1/api/image/caption - Generate image caption
• POST /v1/api/image/prompt - Convert image to prompt

Configuration

Output Formats

• webp (default) - High quality, small size
• png - Lossless quality
• jpg - Standard compressed
• pil - PIL Image object

Available Models & Presets

# Get available options
models = atelier.list_atr_models # Base models
sizes = atelier.list_atr_size # Image sizes
styles = atelier.list_sty_styles # Style presets
lora_models = {
    'svi': atelier.list_atr_lora_svi, # SVI LoRA models
    'flux': atelier.list_atr_lora_flux, # Flux LoRA models
    'rt': atelier.list_atr_lora_rt # RT LoRA models
}

License

See LICENSE for details.