Transforming Images with the SDXL Prompt-to-Prompt Cognitive Actions

23 Apr 2025
Transforming Images with the SDXL Prompt-to-Prompt Cognitive Actions

In the world of image editing, the ability to manipulate visuals through intuitive prompts opens up endless possibilities for developers and artists alike. The adirik/sdxl-prompt-to-prompt API provides a powerful set of Cognitive Actions that leverage advanced techniques like self-attention and cross-attention to edit images seamlessly. Whether you're looking to replace elements, refine details, or adjust specific features of an image, these pre-built actions simplify the process, allowing you to focus on creativity rather than technical intricacies.

Prerequisites

To start using the Cognitive Actions from the adirik/sdxl-prompt-to-prompt API, you will need:

  • An API key for authentication with the Cognitive Actions platform.
  • Basic knowledge of JSON payloads and HTTP requests.

Authentication typically involves passing your API key in the headers of your request. Here’s a generic structure for how to do that in your API calls.

Cognitive Actions Overview

Edit Image Using SDXL Prompt-to-Prompt

The Edit Image Using SDXL Prompt-to-Prompt action allows you to edit images using the SDXL Prompt-to-Prompt technique. This action facilitates modifications by generating an original image and a modified version based on changes you specify, enabling various editing approaches including Replacement, Refinement, or Re-weighting.

Input: This action requires the following fields based on its schema:

  • originalPrompt (string, required): The prompt used to generate the original image.
    Example: "a pink bear riding a bicycle on the beach"
  • promptEditType (string, required): The type of editing to be performed, with options: Replacement, Refinement, or Re-weight.
    Example: "Replacement"
  • seed (integer, optional): A random seed for generation, with a default range of 0 to 65535.
    Example: 864
  • image (string, optional): URI of an optional input image for initial latent variable retrieval.
  • localEdit (string, optional): Comma-separated list of words specifying the area to be changed.
  • editedPrompt (string, optional): The prompt used for editing the original image.
    Example: "a pink dragon riding a bicycle on the beach"
  • guidanceScale (number, optional): Scale for text guidance, defaulting to 7.5.
  • selfReplaceSteps (number, optional): Fraction of diffusion steps for self-attention replacement, defaulting to 0.4.
  • crossReplaceSteps (number, optional): Fraction of diffusion steps for cross-attention replacement, defaulting to 0.8.
  • numInferenceSteps (integer, optional): Number of diffusion denoising steps used for image generation, defaulting to 50.
  • numInversionSteps (integer, optional): Number of diffusion denoising steps for inversion, defaulting to 50.
  • equalizerWords (string, optional): Comma-separated words for re-weighting.
  • equalizerStrengths (string, optional): Comma-separated strengths corresponding to the equalizer words.

Example Input:

{
  "seed": 864,
  "editedPrompt": "a pink dragon riding a bicycle on the beach",
  "originalPrompt": "a pink bear riding a bicycle on the beach",
  "promptEditType": "Replacement",
  "selfReplaceSteps": 0.4,
  "crossReplaceSteps": 0.8
}

Output: The action typically returns an array of URLs pointing to the generated images:

[
  "https://assets.cognitiveactions.com/invocations/7722cb84-9a08-49b8-ac1e-8848039f79ea/35b0bdb7-59ed-4650-8dfc-35ddd46ddc98.png",
  "https://assets.cognitiveactions.com/invocations/7722cb84-9a08-49b8-ac1e-8848039f79ea/649a54e9-4595-46ad-b840-2c0df987f77c.png"
]

Conceptual Usage Example (Python): Here’s how you might call this action using Python:

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 = "4874bf8d-66dd-4f93-85f8-7468c013754b"  # Action ID for Edit Image Using SDXL Prompt-to-Prompt

# Construct the input payload based on the action's requirements
payload = {
    "seed": 864,
    "editedPrompt": "a pink dragon riding a bicycle on the beach",
    "originalPrompt": "a pink bear riding a bicycle on the beach",
    "promptEditType": "Replacement",
    "selfReplaceSteps": 0.4,
    "crossReplaceSteps": 0.8
}

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 this code snippet, replace the placeholder API key and endpoint with your actual credentials. The action_id is specific to the editing action being invoked. The input payload should match the structure as defined in the action's schema.

Conclusion

The adirik/sdxl-prompt-to-prompt Cognitive Actions empower developers to creatively manipulate images with ease. By leveraging the capabilities of prompt-based editing, you can enhance your applications with sophisticated image processing features. Consider exploring these actions further, experimenting with different prompts and editing types to discover the creative possibilities they unlock!

← Back to All Articles