[FEATURE REQ] Protocol methods for `ImageClient`

[stefanosambruna]

How can we help?

How do I specify the number of images? (https://platform.openai.com/docs/api-reference/images/create#images-create-n)

I have the feeling that the support to gpt-image-1 was added but the interfaces didn't change correspondely. For instance, I don't see all the options and values available for the new model, starting from the number of images, arriving to the quality and size.

[jsquire]

Hi @stefanosambruna. Thanks for reaching out and we regret that you're experiencing difficulties. You're correct in that the REST API spec that the latest release was built from did not contain some of the newer options for image generation. Until they're added, you can use these properties by dropping down to the protocol overload for GenerateImagesAsync. This would look something like:

var client = new ImageClient("gpt-4.1", "<<  YOUR API KEY >>");

var rawResponse = await client.GenerateImagesAsync(
    BinaryContent.Create(BinaryData.FromObjectAsJson(new
    {
        model = "gpt-image-1",
        prompt = "A cute baby sea otter",
        n = 1,
        size = "1024x1024"
    })));

var response = ModelReaderWriter.Read<GeneratedImageCollection>(rawResponse.GetRawResponse().Content);    
Console.WriteLine($"Image URL: {response.First().ImageUri}");

EDIT

I adjusted to use the protocol overload that I missed when originally writing this up. The example has been simplified to remove the policy use in favor of the direct method call.