Unlocking Image Generation with alexgenovese/sdxl-lora Cognitive Actions

24 Apr 2025
Unlocking Image Generation with alexgenovese/sdxl-lora Cognitive Actions

In the realm of AI-driven image processing, the alexgenovese/sdxl-lora Cognitive Actions empower developers to generate stunning visuals through advanced models. These pre-built actions leverage the Realistic Vision XL 4.0 model, allowing seamless integration of LoRA (Low-Rank Adaptation) models for enhanced image generation tasks. Whether for artistic endeavors or practical applications, these actions streamline the process of image creation, making it accessible and efficient.

Prerequisites

To effectively use the Cognitive Actions provided by alexgenovese/sdxl-lora, you will need:

  • An API key for the Cognitive Actions platform to authenticate your requests.
  • Basic knowledge of JSON payload structures and how to make HTTP requests.

Authentication typically involves passing the API key in the request headers as follows:

Authorization: Bearer YOUR_COGNITIVE_ACTIONS_API_KEY

Cognitive Actions Overview

Run Inference on Realistic Vision XL with LoRA

Description:
This action performs inference on images using the Realistic Vision XL 4.0 model, incorporating any LoRA model through user-defined parameters. It supports Mac (M1, M2, M3) and CUDA machines, enabling flexible image generation options, including img2img or inpaint modes. Developers can customize masks, seeds, and scheduling algorithms to refine their results.

Category: image-generation

Input

The input for this action requires several fields, which are detailed below:

  • mask (string, optional): URI to the input mask for inpaint mode. Black areas will be preserved, while white areas will be inpainted.
  • seed (integer, optional): Random seed for the image generation process. Defaults to a randomized seed if left blank.
  • image (string, optional): URI of the input image for img2img or inpaint mode.
  • width (integer, optional): Width of the output image. Default is 1024 pixels.
  • height (integer, optional): Height of the output image. Default is 1024 pixels.
  • prompt (string, required): Text prompt guiding the image generation. It serves as the main input.
  • scheduler (string, optional): Algorithm used for scheduling the denoising steps. Defaults to K_EULER.
  • outputCount (integer, optional): Number of images to generate (between 1 and 4). Default is 1.
  • guidanceScale (number, optional): Scaling factor for classifier-free guidance (1 to 50). Default is 7.5.
  • applyWatermark (boolean, optional): Indicates if a watermark should be applied. Defaults to true.
  • negativePrompt (string, optional): Text prompt describing elements to avoid.
  • promptStrength (number, optional): Influence of the prompt on the image (0 to 1). Default is 0.8.
  • refinementSteps (integer, optional): Number of refinement steps for the base image refiner.
  • refinementStyle (string, optional): Style for refining the image. Default is no_refiner.
  • modelResourceUrl (string, optional): URL to the LoRA model resources.
  • modelScaleFactor (number, optional): Additive scale factor for LoRA models (0 to 1). Default is 0.6.
  • noiseFractionHigh (number, optional): Fraction of noise used in expert ensemble refiner (0 to 1). Default is 0.8.
  • inferenceStepCount (integer, optional): Total number of denoising steps during inference (1 to 500). Default is 50.

Example Input:

{
  "seed": 100,
  "width": 1024,
  "height": 1024,
  "prompt": "one violet bag aeqq on the table with lights from the top in a luxury suite of hotel",
  "scheduler": "K_EULER",
  "outputCount": 1,
  "guidanceScale": 7.5,
  "applyWatermark": true,
  "negativePrompt": "Asian, cartoon, 3d, (disfigured), (bad art), (deformed), (poorly drawn), (extra limbs), (close up), strange colors, blurry, boring, sketch, lackluster, big breast, large breast, huge breasts, face portrait, self-portrait, signature, letters, watermark, disfigured, kitsch, ugly, oversaturated, greain, low-res, deformed, blurry, bad anatomy, poorly drawn face, mutation, mutated, extra limb, poorly drawn hands, missing limb, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, disgusting, poorly drawn, childish, mutilated, mangled, old, surreal, calligraphy, sign, writing, watermark, text, body out of frame, extra legs, extra arms, extra feet, out of frame, poorly drawn feet, cross-eye",
  "promptStrength": 0.8,
  "refinementStyle": "no_refiner",
  "modelResourceUrl": "https://pbxt.replicate.delivery/dgeTJ4xBtfqZj0wGnbsoIMSm9sPI5lDzzeyxeQTmCtIeI17LC/trained_model.tar",
  "modelScaleFactor": 0.8,
  "noiseFractionHigh": 0.8,
  "inferenceStepCount": 50
}

Output

The action typically returns a list of image URLs generated based on the input parameters.

Example Output:

[
  "https://assets.cognitiveactions.com/invocations/68de152f-e883-41d1-84ab-9bf3e5bcbd79/51a7f1d2-8eaf-4f6f-b452-3f1a343d1f29.png"
]

Conceptual Usage Example (Python)

Below is a conceptual Python script illustrating how to call the Cognitive Actions execution endpoint:

import requests
import json

# Replace with your Cognitive Actions API key and endpoint
COGNITIVE_ACTIONS_API_KEY = "YOUR_COGNITIVE_ACTIONS_API_KEY"
COGNITIVE_ACTIONS_EXECUTE_URL = "https://api.cognitiveactions.com/actions/execute"  # Hypothetical endpoint

action_id = "d0580647-ba7f-437b-a249-50003140b148"  # Action ID for Run Inference on Realistic Vision XL with LoRA

# Construct the input payload based on the action's requirements
payload = {
    "seed": 100,
    "width": 1024,
    "height": 1024,
    "prompt": "one violet bag aeqq on the table with lights from the top in a luxury suite of hotel",
    "scheduler": "K_EULER",
    "outputCount": 1,
    "guidanceScale": 7.5,
    "applyWatermark": True,
    "negativePrompt": "Asian, cartoon, 3d, (disfigured), (bad art), (deformed), (poorly drawn), (extra limbs), (close up), strange colors, blurry, boring, sketch, lackluster, big breast, large breast, huge breasts, face portrait, self-portrait, signature, letters, watermark, disfigured, kitsch, ugly, oversaturated, greain, low-res, deformed, blurry, bad anatomy, poorly drawn face, mutation, mutated, extra limb, poorly drawn hands, missing limb, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, disgusting, poorly drawn, childish, mutilated, mangled, old, surreal, calligraphy, sign, writing, watermark, text, body out of frame, extra legs, extra arms, extra feet, out of frame, poorly drawn feet, cross-eye",
    "promptStrength": 0.8,
    "refinementStyle": "no_refiner",
    "modelResourceUrl": "https://pbxt.replicate.delivery/dgeTJ4xBtfqZj0wGnbsoIMSm9sPI5lDzzeyxeQTmCtIeI17LC/trained_model.tar",
    "modelScaleFactor": 0.8,
    "noiseFractionHigh": 0.8,
    "inferenceStepCount": 50
}

headers = {
    "Authorization": f"Bearer {COGNITIVE_ACTIONS_API_KEY}",
    "Content-Type": "application/json"
}

try:
    response = requests.post(
        COGNITIVE_ACTIONS_EXECUTE_URL,
        headers=headers,
        json={"action_id": action_id, "inputs": payload}  # Hypothetical structure
    )
    response.raise_for_status()  # Raise an exception for bad status codes (4xx or 5xx)

    result = response.json()
    print("Action executed successfully:")
    print(json.dumps(result, indent=2))

except requests.exceptions.RequestException as e:
    print(f"Error executing action {action_id}: {e}")
    if e.response is not None:
        print(f"Response status: {e.response.status_code}")
        try:
            print(f"Response body: {e.response.json()}")
        except json.JSONDecodeError:
            print(f"Response body: {e.response.text}")

In the code snippet above:

  • Replace YOUR_COGNITIVE_ACTIONS_API_KEY with your actual API key.
  • The payload is constructed based on the required input fields for the action.
  • The script executes a POST request to the hypothetical Cognitive Actions endpoint and handles potential errors gracefully.

Conclusion

The alexgenovese/sdxl-lora Cognitive Actions provide developers with powerful tools for image generation, enabling creative possibilities and practical applications alike. By harnessing the capabilities of the Realistic Vision XL model with LoRA enhancements, you can easily integrate advanced image processing into your applications. Explore the full potential of these actions and consider how they can elevate your projects to new visual heights!

← Back to All Articles