Qubico Logo

Documentation

Advanced Features

LoRA & ControlNet

Professional styling and precision control for advanced image generation workflows.

LoRA Models
Low-Rank Adaptation for custom styles and subjects
  • • Custom artistic styles
  • • Character consistency
  • • Brand-specific aesthetics
  • • Fine-tuned adaptations
ControlNet
Precise structural and compositional control
  • • Pose and gesture control
  • • Depth and perspective guidance
  • • Edge and line detection
  • • Architectural precision
Text-to-Image with LoRA
Generate images from text prompts enhanced with custom LoRA models

Request Example

POST https://api.qubico.ai/v1/inference
Content-Type: application/json
x-api-key: YOUR_API_KEY

{
  "model": "Qubico/flux1-dev-advanced",
  "task_type": "txt2img-lora",
  "input": {
    "prompt": "person enjoying a day at the park, full hd, cinematic lighting",
    "negative_prompt": "low quality, ugly, distorted, artifacts",
    "guidance_scale": 4.0,
    "lora_settings": [
      {
        "lora_type": "lora-dragon",
        "lora_strength": 0.7
      }
    ],
    "width": 1024,
    "height": 1024
  }
}
Parameter Reference
Complete parameter documentation for LoRA and ControlNet features

Required Parameters

lora_type
string
required

LoRA model identifier. See available models for complete list.

Optional Parameters

lora_strength
float
optional

LoRA influence strength. Range: 0.0-2.0. Default: 1.0. Higher values = stronger style effect. Recommended: 0.7-0.8 for balanced results.

Available Models & Controls
Complete reference for all LoRA models and ControlNet types

View detailed information about all available LoRA models and ControlNet types, including examples, use cases, and recommended settings.

View Available Models & Controls →
Best Practices
Tips for optimal results with LoRA and ControlNet

LoRA Usage

  • • Start with strength 0.7-0.8 for balanced results
  • • Use multiple LoRAs sparingly (max 2-3)
  • • Lower strength for subtle style hints
  • • Higher strength for dramatic transformations
  • • Test different combinations for unique styles

ControlNet Usage

  • • Use high-contrast control images for better results
  • • Adjust control_strength based on desired precision
  • • Combine with appropriate prompts for context
  • • Preview preprocessed images to verify detection
  • • Use lower denoise values for stronger control
Pricing & Credits
Credit costs for LoRA and ControlNet features

LoRA Enhanced

4.0
credits per image

ControlNet

5.0
credits per image

ControlNet + LoRA

6.0
credits per image