Fine-Tuning Your Image Generation with the FLUX.1-dev Cognitive Actions

23 Apr 2025
Fine-Tuning Your Image Generation with the FLUX.1-dev Cognitive Actions

In the world of AI-driven image generation, the ostris/flux-dev-lora-trainer API offers powerful Cognitive Actions that enable developers to fine-tune models for creative applications. With the Fine-Tune FLUX.1-dev action, you can generate images tailored to specific prompts and configurations. This action supports img2img and inpainting modes, allowing for customizable output that meets your project needs.

Prerequisites

Before diving into the integration, ensure you have the following:

  • An API key for accessing the Cognitive Actions platform.
  • Basic familiarity with making HTTP requests and handling JSON payloads.

Authentication typically involves passing your API key in the request headers to authorize access to the Cognitive Actions API.

Cognitive Actions Overview

Fine-Tune FLUX.1-dev

The Fine-Tune FLUX.1-dev action is designed to fine-tune the FLUX.1-dev model, enabling the generation of images based on specific prompts. This action is part of the image-generation category and allows for intricate control over the generated visuals, supporting various configurations and adjustments to achieve high-quality outputs.

Input

The input for this action requires a JSON object that adheres to the following schema:

{
  "prompt": "your_prompt_here",
  "model": "dev",
  "loraScale": 1,
  "guidanceScale": 3.5,
  "numberOfOutputs": 1,
  "imageAspectRatio": "1:1",
  "imageOutputFormat": "webp",
  "imageOutputQuality": 80,
  "numberOfInferenceSteps": 28
}

Required Fields:

  • prompt: The text prompt for the image generation. For example, "puppy".

Optional Fields:

  • mask: URI for the input mask in inpainting mode.
  • seed: Integer for reproducibility.
  • image: URI for the input image in img2img or inpainting mode.
  • model: Select either "dev" or "schnell".
  • width: Desired image width (if using custom aspect ratio).
  • height: Desired image height (if using custom aspect ratio).
  • loraWeights: Custom LoRA weights.
  • additionalLora: Combine with another LoRA model.
  • numberOfOutputs: Specify how many images to generate (1-4).
  • imageAspectRatio: Aspect ratio for the generated image.
  • imageOutputFormat: Output format (webp, jpg, png).
  • imageOutputQuality: Quality of the output image (0-100).
  • numberOfInferenceSteps: Number of steps for inference.

Example Input:

{
  "model": "dev",
  "prompt": "puppy",
  "loraScale": 1,
  "guidanceScale": 3.5,
  "numberOfOutputs": 1,
  "imageAspectRatio": "1:1",
  "imageOutputFormat": "webp",
  "imageOutputQuality": 80,
  "numberOfInferenceSteps": 28
}

Output

When the action is executed, it typically returns a JSON object containing links to the generated images. Here’s an example of the output you can expect:

[
  "https://assets.cognitiveactions.com/invocations/a7ec0c75-dde5-4bdf-97d4-e448821fc693/cebae1e2-63af-4e7d-b965-08729addca37.webp"
]

Conceptual Usage Example (Python)

Below is a conceptual Python snippet demonstrating how to call the Fine-Tune FLUX.1-dev action:

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 = "271dff44-8ef2-4095-8bf6-e8ce5de6bbda" # Action ID for Fine-Tune FLUX.1-dev

# Construct the input payload based on the action's requirements
payload = {
    "model": "dev",
    "prompt": "puppy",
    "loraScale": 1,
    "guidanceScale": 3.5,
    "numberOfOutputs": 1,
    "imageAspectRatio": "1:1",
    "imageOutputFormat": "webp",
    "imageOutputQuality": 80,
    "numberOfInferenceSteps": 28
}

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}")

This code snippet shows how to structure the input payload for the Fine-Tune FLUX.1-dev action and handle the response. Remember, the endpoint URL and request structure are illustrative.

Conclusion

The Fine-Tune FLUX.1-dev action provides developers with powerful capabilities for generating and fine-tuning images based on specific prompts. With customizable settings for model selection, output format, and quality, you can create unique visuals that meet your application's needs. Consider experimenting with different configurations to explore the full potential of this action in your projects!

← Back to All Articles