---

name: z-image-turbo-controlnet-lora description: "Z Image | Turbo | Controlnet | Lora. Generate images with ControlNet guidance and LoRA weights. Triggers: controlnet, lora, guided generation, z image, pose to image, depth" allowed-tools: Bash(curl *), WebFetch

Z Image | Turbo | Controlnet | Lora

Generate images with ControlNet-guided generation and LoRA weight support. Use canny edges, depth maps, or pose detection from reference images combined with custom LoRA models for precise control over composition and style.

Quick Start

Requires an each::labs API key. Get one at eachlabs.ai.

curl -X POST https://api.eachlabs.ai/v1/prediction \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "model": "z-image-turbo-controlnet-lora",
    "version": "0.0.1",
    "input": {
      "prompt": "A futuristic robot in the same pose, metallic chrome finish",
      "image_url": "https://example.com/person-pose.jpg",
      "preprocess": "pose",
      "loras": [{"path": "https://example.com/scifi-lora.safetensors", "scale": 0.8}],
      "control_scale": 0.9,
      "image_size": "auto",
      "output_format": "png"
    }
  }'

Parameters

Parameter	Type	Default	Description
acceleration	string	none	The acceleration level to use. Options: none, regular, high
control_end	number	0.4	The end of the controlnet conditioning.
control_scale	number	0.9	The scale of the controlnet conditioning.
control_start	number	0	The start of the controlnet conditioning.
enable_prompt_expansion	boolean	false	Whether to enable prompt expansion.
enable_safety_checker	boolean	true	If set to true, the safety checker will be enabled.
image_size	string	auto	The size of the generated image. Options: square_hd, square, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9, auto
image_url	string	-	URL of Image for ControlNet generation.
loras	array	-	List of LoRA weights to apply.
num_images	integer	1	The number of images to generate.
num_inference_steps	integer	8	The number of inference steps to perform.
output_format	string	png	The format of the generated image. Options: jpeg, png, webp
preprocess	string	none	Preprocessing to apply to the image. Options: none, canny, depth, pose
prompt	string	-	The prompt to generate an image from.
seed	integer	-	Seed for reproducible generation.

Examples

Canny edge-guided generation:

curl -X POST https://api.eachlabs.ai/v1/prediction \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "model": "z-image-turbo-controlnet-lora",
    "version": "0.0.1",
    "input": {
      "prompt": "Elegant architectural rendering, glass and steel",
      "image_url": "https://example.com/building-sketch.jpg",
      "preprocess": "canny",
      "loras": [{"path": "https://example.com/architecture-lora.safetensors", "scale": 0.7}]
    }
  }'

Depth-guided with fine-tuned control:

curl -X POST https://api.eachlabs.ai/v1/prediction \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "model": "z-image-turbo-controlnet-lora",
    "version": "0.0.1",
    "input": {
      "prompt": "An enchanted forest scene with magical creatures",
      "image_url": "https://example.com/forest-photo.jpg",
      "preprocess": "depth",
      "loras": [{"path": "https://example.com/fantasy-lora.safetensors", "scale": 0.9}],
      "control_scale": 0.7,
      "control_start": 0,
      "control_end": 0.6,
      "acceleration": "regular",
      "num_images": 2,
      "seed": 42
    }
  }'

Related Models

• Z Image | Turbo | Lora - Text-to-image with LoRA
• Z Image | Turbo | Image to Image | Lora - Image-to-image with LoRA
• Z Image | Turbo | Controlnet - ControlNet without LoRA

Documentation

• each::labs Docs
• API Reference