gpt-image-1 using openai SDK?

Question

gpt-image-1 using openai SDK?

Amit Gelber 0

Hi,
How are we expected to work with GPT-Image-1 with Azure?

seems there is no support on AzureOpenAI SDK, and there are auth issues with using OpenAI SDK directly.

Please share examples.

Thank you.

SRILAKSHMI C 11,215 Reputation points Microsoft External Staff Moderator

2025-05-29T14:53:02.0133333+00:00

Hi @Amit Gelber,

Following up to see if the below answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank you!
SRILAKSHMI C 11,215 Reputation points Microsoft External Staff Moderator

2025-05-30T05:30:22.5533333+00:00

Hi @Amit Gelber,

Just checking in to see if you have got a chance to see my response to your question in resolving the issue.

If you are still facing any further issues, please don't hesitate to reach out to us. We are happy to assist you.

Looking forward to your response and appreciate your time on this.

If you feel that your quires have been resolved, please accept the answer by clicking the "Upvote" and "Accept Answer" on the post.

Thank you!

3 answers

Your answer

SRILAKSHMI C 11,215 Reputation points Microsoft External Staff Moderator

2025-05-29T14:53:02.0133333+00:00

Hi @Amit Gelber,

Following up to see if the below answer was helpful. If this answers your query, do click Accept Answer and Yes for was this answer helpful. And, if you have any further query do let us know.

Thank you!
SRILAKSHMI C 11,215 Reputation points Microsoft External Staff Moderator

2025-05-30T05:30:22.5533333+00:00

Hi @Amit Gelber,

Just checking in to see if you have got a chance to see my response to your question in resolving the issue.

If you are still facing any further issues, please don't hesitate to reach out to us. We are happy to assist you.

Looking forward to your response and appreciate your time on this.

If you feel that your quires have been resolved, please accept the answer by clicking the "Upvote" and "Accept Answer" on the post.

Thank you!

Answer 1

Hello @Amit Gelber,

If you're trying to work with image inputs using GPT-Image-1 (also referred to as gpt-4o or gpt-4-vision-preview) on Azure, it's important to know that Azure OpenAI Service does not currently support vision models. This includes both the Azure OpenAI SDK and the REST API endpoints exposed by Azure. This limitation is the primary reason you're encountering issues when trying to use these models.

You might have attempted to use the OpenAI SDK (available from OpenAI directly) with your Azure OpenAI resource. Unfortunately, this won’t work due to authentication differences. The OpenAI SDK requires OpenAI-issued API keys, which are not interchangeable with Azure resource keys. If you try to use an Azure key with the OpenAI SDK, you’ll get authentication errors. This is a known limitation and expected behavior.

If you're facing auth issues using the OpenAI SDK directly, make sure that you've correctly set up your Azure OpenAI resource and that you are using the right keys. Ensure that you’ve configured the signed headers correctly if applicable.
The Azure OpenAI SDK currently does not support any models with image input capabilities, such as gpt-4o. Even though gpt-4o is in public preview, the image input feature is not yet available on Azure.

At this time, Microsoft has not announced a specific timeline for when vision-enabled models will be available through the Azure OpenAI Service. You can monitor their progress and announcements via the Azure OpenAI release notes and the Azure Updates portal.

Practical examples are available on the Azure SDK for [JavaScript GitHub repository], which can help you understand how to implement the functionality you require.

I Hope this helps. Do let me know if you have any further queries.

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thank you!

Answer 2

Sina Salam 26,666 Volunteer Moderator

Hello SriLakshmi C,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that you are having issue with your gpt-image-1 using openai SDK.

You can use GPT-Image-1 - GPT-4o on Azure via Azure AI Studio or REST API. SDKs like openai are not compatible with Azure due to auth differences. Use REST as shown below or try the Chat Playground in Azure AI Studio - https://oai.azure.com. For automation, follow the GitHub sample project - https://github.com/Azure-Samples/openai-chat-vision-quickstart and ensure your model is correctly deployed and accessible. And for API Version Reference - https://learn.microsoft.com/en-us/azure/ai-services/openai/reference

A. Use Azure AI Studio - Low-Code

Go to: https://oai.azure.com/
Use "Chat Playground" and choose:
- Model: GPT-4o or GPT-4 Turbo with Vision
- Deployment: Ensure it's enabled in your Azure OpenAI resource
- Upload an image and chat interactively

Use Azure REST API - This is a python example;Use Azure REST API - This is a python example;

   import requests
   import base64
   # Load and encode the image
   with open - "image.jpg", "rb" as image_file:
   base64_image = base64.b64encode - image_file.read - .decode - "utf-8"
   # Replace with your Azure details
   endpoint = "https://<your-resource-name>.openai.azure.com/openai/deployments/<deployment-name>/chat/completions?api-version=2024-02-15-preview"
   api_key = "<your-azure-api-key>"
   headers = {
   "Content-Type": "application/json",
   "api-key": api_key
   }
   data = {
   "messages": 
   {"role": "system", "content": "You are a helpful assistant."},
   {
       "role": "user",
       "content": 
           {"type": "text", "text": "What’s in this image?"},
           {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{base64_image}"}}
       
   }
   ,
   "max_tokens": 1000
   }
   response = requests.post - endpoint, headers=headers, json=data
   print - response.json

I hope this is helpful! Do not hesitate to let me know if you have any other questions or clarifications.

Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Almog Ben Simon 0 Reputation points

2025-05-29T10:58:00.7666667+00:00

Hi, thanks for the response. So, if I understand correctly, I can use the gpt-image-1 model for image generation with input images via gpt-4o and the image generation tooling on Azure. However, I can't access gpt-image-1 directly for this functionality—is that right?
Sina Salam 26,666 Reputation points Volunteer Moderator

2025-05-29T12:38:59.5+00:00
Hello Almog Ben Simon,

Thank you for clarification and your feedback.

You cannot directly access gpt-image-1 for image generation via Azure or the OpenAI SDK.

For the same purpose of gpt-image-1 you can use GPT-4o (or GPT-4 Turbo with Vision) on Azure to analyze images (e.g., describe, interpret, or answer questions about them).

For image generation, Azure provides DALL·E models, which are separate from GPT-4o or gpt-image-1.

If your goal is:

Image understanding (e.g., “What’s in this image?”): Use GPT-4o via Azure REST API or Chat Playground.

Image generation (e.g., “Create an image of a cat on Mars”): Use DALL·E via Azure OpenAI Studio or REST API.

Success.
Amit Gelber 0 Reputation points

2025-05-29T13:16:34.06+00:00

Hi, what about masking? Provide an image and ask for minor changes?
What are the options for these capabilities: https://www.datacamp.com/tutorial/gpt-image-1
specifically Editing Images with GPT-Image-1 and Using masks
Sina Salam 26,666 Reputation points Volunteer Moderator

2025-05-29T14:06:06.2466667+00:00
Hi, what about masking? Provide an image and ask for minor changes? What are the options for these capabilities: https://www.datacamp.com/tutorial/gpt-image-1 specifically Editing Images with GPT-Image-1 and Using masks

As you have seen in the tutorial, if it works for you fine. That might be a nice workaround.

On Azure:

Azure currently supports image editing with masking through a combination of services, even though GPT-Image-1 is not directly accessible for image generation or editing via Azure. But you can achieve similar functionality using GPT-4o (or GPT-4 Turbo with Vision) to:

Analyze an image.

Identify the region to be edited.

Generate a mask (e.g., a polygon or binary mask) based on a natural language prompt.

For an example, “Generate a mask for the TV and the cabinet on the right side of this 720x720 image.” GPT-4o can return polygon coordinates or a NumPy array representing the mask.

Also, once you have the mask, you can use Stable Diffusion Inpainting models (like runwayml-stable-diffusion-inpainting) deployed via Azure Machine Learning (AzureML) to:

Apply the mask.

Use a prompt to describe the desired change.

Generate the edited image.

For an example: You can “Replace the TV with a bookshelf.”

You just need to work more on your Workflow, get the right Tools & Resources, you can also read more here:

https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/dall-e

https://techcommunity.microsoft.com/blog/machinelearningblog/image-editing-using-azure-openai-and-azure-machine-learning/4339077

Answer 3

Amit Gelber 0

We're primarily interested in GPT-image-1's image capabilities for editing purposes, both with and without masking. While I understand that GPT-4o supports some level of image generation, there's clearly a reason why OpenAI and Microsoft introduced a dedicated model like GPT-image-1.

From what I gather in this thread, GPT-image-1 with with input image support is currently only accessible via OpenAI directly, and not yet available through Azure — is that correct?

Share via

gpt-image-1 using openai SDK?

3 answers

Your answer