[Feature Request] Add Image-to-Image (img2img) Code Sample for FLUX.2-pro on Azure AI Foundry

[BeastxD7]

Description

There is currently no code sample or documentation for the image-to-image (img2img) use case with the FLUX.2-pro model deployed on Azure AI Foundry.

The only available samples demonstrate text-to-image generation (/openai/deployments/{deployment}/images/generations), but there is no guidance on how to perform image editing / image variation — where the input is both an image AND a text prompt.

Current Sample Code in AI Foundry

Expected Behavior

Azure AI Foundry should provide:

1. A code sample (Python) demonstrating how to send an input image along with a text prompt to the FLUX.2-pro deployment to generate a modified/edited version of the input image.
2. REST API documentation specifying:
   • The correct endpoint path (e.g., /images/edits or a parameter in /images/generations)
   • How to pass the input image (base64 in JSON body, multipart form data, image URL, etc.)
   • Supported API versions for this feature
   • Any model-specific parameters for controlling edit strength, guidance scale, etc.

Current Behavior

• The FLUX.2-pro model is deployed successfully as a GlobalStandard deployment on Azure AI Foundry.
• Text-to-image generation works via POST /openai/deployments/FLUX.2-pro/images/generations.
• Attempting to use /openai/deployments/FLUX.2-pro/images/edits (similar to DALL-E edit API) returns 404 Resource not found.
• There is no documented way to include an input/reference image in the request payload.

Steps to Reproduce

1. Deploy FLUX.2-pro model on Azure AI Foundry (GlobalStandard deployment).
2. Attempt to call the image edit endpoint:

import httpx

endpoint = "https://<resource>.cognitiveservices.azure.com"
deployment = "FLUX.2-pro"
api_key = "<your-api-key>"

# Attempt 1: /images/edits endpoint (returns 404)
url = f"{endpoint}/openai/deployments/{deployment}/images/edits?api-version=2024-05-01-preview"

headers = {"api-key": api_key}
files = {"image": ("input.png", image_bytes, "image/png")}
data = {"prompt": "Change the background to a sunset", "n": "1", "size": "1024x1024"}

response = httpx.post(url, data=data, files=files, headers=headers)
print(response.status_code)  # 404

# Attempt 2: /images/generations with image in body (no documented parameter for input image)
url = f"{endpoint}/openai/deployments/{deployment}/images/generations?api-version=2024-05-01-preview"

headers = {"Content-Type": "application/json", "api-key": api_key}
payload = {
    "prompt": "Change the background to a sunset",
    "n": 1,
    "size": "1024x1024",
    # How to include the input image? No documentation available.
}

response = httpx.post(url, json=payload, headers=headers)
# Works but generates from scratch — doesn't use input image

Environment

• Model: FLUX.2-pro (v1, by Black Forest Labs)
• Deployment type: GlobalStandard
• Resource type: Azure AI Foundry (cognitiveservices.azure.com)
• API versions tested: 2024-05-01-preview, 2025-04-01-preview

Suggested Solution

1. If img2img is supported: Add a code sample to the [Azure AI Foundry SDK samples Code] showing how to pass an input image to FLUX.2-pro for editing/variation.
2. If img2img is NOT yet supported: Document this limitation clearly in the FLUX.2-pro model card on Azure AI Foundry, and provide an ETA or roadmap for when image-to-image support will be available.

Additional Context

FLUX.2-pro natively supports image-to-image generation (as documented on Black Forest Labs). The Azure deployment should ideally expose this capability through the REST API.