Calling the OpenAI API to create Dall-E images from C#

This page has 0 threads | Add post

Blogs

Finding our blogs useful? Please consider helping us to share more!

Calling the OpenAI API to create Dall-E images from C#
For our newsletter competition this month everyone gets one shot only at creating a perfect image in Dall-E. This blog takes a behind-the-scenes look to explain the nuts and bolts of how we used the OpenAI API to get this to work.

Posted by Andy Brown on 09 December 2024

In this blog

Getting started with the OpenAI API
Adding the OpenAI Nuget Package
Starting a new API request
Generating your image
Saving the image
Some final thoughts

For our December 2024 newsletter competition we invited people to create a picture using ChatGPT:

I'm writing this blog in real time, so don't yet know what this picture will look like! I'm not optimistic though ...

When someone clicks on the button to create an image, a call goes out to the OpenAI API. This then sends back the image:

I was right to be pessimistic.

The question is: how did this work?

This doesn't claim to be a tutorial, but instead gives you the steps that you need to follow in outline. It will be especially useful for anyone who is trying to call the OpenAI API from C#.

Getting started with the OpenAI API

To be able to interact with ChatGPT programmatically you'll need to use the OpenAI API (Application Programming Interface) - and for that you'll need an API key. You can get this by signing up here:

Once you've signed up you can click on this link to see information to do with your account.

You can use the dashboard (among other things) to see how much your account is costing you:

Our experience so far is that generating an image costs a matter of cents.

Adding the OpenAI Nuget Package

If you want to be able to use C# to call the OpenAI API you shouuld add the OpenAI package using Nuget:

Add this package using the NuGet package manager in Visual Studio. Notice that it's still officially in beta.

Starting a new API request

When someone requests a picture you should initiate a new API request (obviously the API key will be the one given when you first created your API account):

// authenticate yourself

var openAiApiKey = "sk-proj-long-sequence-of-meaningless-characters";

// would be better to use an environment variable

// Environment.GetEnvironmentVariable("OPENAI_API_KEY");

// say which model you are interacting with

OpenAI.Images.ImageClient imageClient = new OpenAI.Images.ImageClient(

model: "dall-e-3",

apiKey: openAiApiKey

);

When creating a new image you can choose which model to use - here we're using version 3 of Dall-E (the latest one).

Generating your image

Having started an API request like this, you can now give some more details on the image you want to create and then request it:

// the image generation options

var igo = new ImageGenerationOptions();

igo.Size = new OpenAI.Images.GeneratedImageSize(1024, 1024);

igo.Quality = OpenAI.Images.GeneratedImageQuality.Standard;

// generate the image requested

var image = imageClient.GenerateImage(

prompt: Prompt,

options: igo

);

If you've some experience of coding in C# it's reasonably easy to see what this code is doing, but I found it hard to find any documentation online about the structure of the OpenAI add-in's namespaces.

Saving the image

The OpenAI API will store images for a couple of hours only, so the onus is on you now to save the image you've created. To do this, first get its website address:

// get the URL to this image

string imageUrl = image.Value.ImageUri.AbsoluteUri;

You can now save this image. One way to do this - and I confess I got ChatGPT to recommend this - is shown below:

// images last for about 2 hours, so save this

using (HttpClient client = new HttpClient())

{

// need to call async method to get image

HttpResponseMessage response = client.GetAsync(imageUrl).Result;

if (response.IsSuccessStatusCode)

{

byte[] imageBytes = response.Content.ReadAsByteArrayAsync().Result;

string filePath = Server.MapPath(@"~/special/customimages/");

string fileName = $"i{customImageRecord.SpecialCreateImageId.ToString()}.webp";

var imageFolder = new System.IO.DirectoryInfo(filePath);

System.IO.File.WriteAllBytes(filePath + fileName, imageBytes);

}

Notice that you have to use an async method to get the image from the website. Obviously the lines setting the file path and file name will be different for your application, but everything else could stay the same.

Some final thoughts

The above code is as basic as I could make it. Here are two obvious improvements you could make:

Improvement	Explanation
Add error-trapping	What happens if your users request more than 5 images in a minute, taking you over the API limit? Or if their prompts make no sense, or violate decency, or (worse still) copyright?
Use async calls	In my case I wanted the screen to pause until an image was returned, but you might want to use async calls in your controller to allow other tasks to run at the same time.

If you've read all of this thinking "but why didn't he do this in Python?", the answer is that I did; but I couldn't find an easy way to call a Python script from within ASP.NET MVC!

Some other pages relevant to the above blog include:

Artificial Intelligence training

This blog has 0 threads

Add a new post

BLOGS BY TOPIC

BLOGS BY AUTHOR

Getting started with the OpenAI API

Adding the OpenAI Nuget Package

Starting a new API request

Generating your image

Saving the image

Some final thoughts

Head office

London

Manchester