---

name: instant-id description: "Instant ID Generate Avatar | AI Avatar Generation. Generate personalized avatars from face photos with extensive controlnet and model options. Triggers: instant id, avatar generator, face avatar, id avatar, personalized avatar" allowed-tools: Bash(curl *), WebFetch

Instant ID Generate Avatar

Generate personalized avatars from face photos with extensive ControlNet options (pose, canny, depth), multiple SDXL base models, and configurable identity preservation strength.

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": "instant-id",
    "version": "0.0.1",
    "input": {
      "image": "https://example.com/face-photo.jpg",
      "prompt": "a person as a fantasy warrior in golden armor, epic background",
      "num_inference_steps": 4,
      "guidance_scale": 7.5,
      "controlnet_conditioning_scale": 0.8,
      "ip_adapter_scale": 0.8
    }
  }'

Parameters

Parameter	Type	Default	Description
canny_strength	number	0.3	Canny ControlNet strength (only if enable_canny_controlnet is true)
controlnet_conditioning_scale	number	0.8	Scale for IdentityNet strength (for fidelity)
depth_strength	number	0.5	Depth ControlNet strength (only if enable_depth_controlnet is true)
disable_safety_checker	boolean	false	Disable safety checker
enable_canny_controlnet	boolean	false	Enable canny edge detection control
enable_depth_controlnet	boolean	false	Enable depth control
enable_lcm	boolean	false	Enable LCM for faster inference
enable_pose_controlnet	boolean	true	Enable Openpose ControlNet
enhance_nonface_region	boolean	true	Enhance non-face region
guidance_scale	number	7.5	Scale for classifier-free guidance
height	integer	640	Height of output image
image	string	-	Input face image
ip_adapter_scale	number	0.8	Scale for image adapter strength (for detail)
lcm_guidance_scale	number	1.5	Guidance scale when using LCM
lcm_num_inference_steps	integer	5	Steps when using LCM
negative_prompt	string	-	Negative prompt
num_inference_steps	integer	4	Number of denoising steps
pose_image	string	-	Optional reference pose image
pose_strength	number	0.4	Openpose ControlNet strength
prompt	string	a person	Input prompt
scheduler	string	EulerDiscreteScheduler	Scheduler. enum: DEISMultistepScheduler, HeunDiscreteScheduler, EulerDiscreteScheduler, DPMSolverMultistepScheduler, DPMSolverMultistepScheduler-Karras, DPMSolverMultistepScheduler-Karras-SDE
sdxl_weights	string	stable-diffusion-xl-base-1.0	Base model. enum: stable-diffusion-xl-base-1.0, juggernaut-xl-v8, afrodite-xl-v2, albedobase-xl-20, albedobase-xl-v13, animagine-xl-30, anime-art-diffusion-xl, anime-illust-diffusion-xl, dreamshaper-xl, dynavision-xl-v0610, guofeng4-xl, nightvision-xl-0791, omnigen-xl, pony-diffusion-v6-xl, protovision-xl-high-fidel
seed	integer	-	Random seed
width	integer	640	Width of output image

Examples

Anime avatar with custom base model:

curl -X POST https://api.eachlabs.ai/v1/prediction \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "model": "instant-id",
    "version": "0.0.1",
    "input": {
      "image": "https://example.com/selfie.jpg",
      "prompt": "a person as an anime character, colorful, detailed, vibrant background",
      "sdxl_weights": "animagine-xl-30",
      "controlnet_conditioning_scale": 0.7,
      "ip_adapter_scale": 0.9,
      "num_inference_steps": 6,
      "guidance_scale": 8,
      "width": 640,
      "height": 640
    }
  }'

Pose-guided avatar with depth:

curl -X POST https://api.eachlabs.ai/v1/prediction \
  -H "Content-Type: application/json" \
  -H "X-API-Key: $EACHLABS_API_KEY" \
  -d '{
    "model": "instant-id",
    "version": "0.0.1",
    "input": {
      "image": "https://example.com/face.jpg",
      "pose_image": "https://example.com/pose-reference.jpg",
      "prompt": "a person as a cyberpunk hacker in neon-lit room, detailed, high quality",
      "enable_pose_controlnet": true,
      "enable_depth_controlnet": true,
      "pose_strength": 0.5,
      "depth_strength": 0.4,
      "guidance_scale": 7.5,
      "num_inference_steps": 5,
      "seed": 42
    }
  }'

Related Models

• instant-id-ip-adapter - Styled avatar generation (3D, Emoji, Pixels)
• photomaker - Identity-preserving photo generation
• become-image - Style transfer with identity preservation

Documentation

• each::labs Docs
• API Reference