How To Use DALL-E 2 To Generate AI Images – Easy Steps

How to Use DALL-E 2? DALL·E 2 started the ai image generator frenzy when it was launched in 2022 by OpenAi. Dalle2 is still one of the best ai image generators available on the Internet. DallE 2 AI is among the top Ai image generators along with the likes of Stable Diffusion, Midjourney, and Imagen by Google.

Dalle 2 AI originally birthed the idea of generating images from giving text prompt to a software. For example: “A cow in a space ship” and the dalle2 will generate an image showing exactly that. When it first came out, it became popular because of its wow factor and people used to generate exciting and erratic results.

In september 2023, OpenAi launched DALL·E 3 which was even more powerful generative ai model for generating images from text prompts. DALL·E 3 is available to all ChatGPT Plus and Enterprise users, and its API access will be available be available soo and in Labs later this fall according to the OpenAi Website.

The technology behind generative ai that generates images from text is a decade old. But the powerful GPUs (hardware) and vast training data has made it better. For now, Dall E2 is no replacement for photography or photo editing. We are seeing use cases of ai generated images in eCommerce, Advertising, print media, entertainment, and for a variety of consumer apps made mostly for fun and creating interesting content.

As the generative ai advances, it may eventually make traditional photography and photo editing obsolete. Till then, we can understand how it works, what are the use cases of Dall E 2 and how we can use and benefit from them. So let’s dive in:

What is DALL·E 2?

DALL·E 2 is a machine learning model (or an AI system) developed by OpenAI (the creators of ChatGPT) that generates realistic images from text prompts. Dall E is trained on a dataset “WebImageText,” which has 400 million pairs of image and captions. The total number of words in WebImageText is almost same as WebText, which has around 40 GB of text data.

OpenAI’s DALL·E 2 has 12-billion parameters that powers its Text-to-image Generation capabilities. DALL-E2, a 12-billion parameter image-to-text generation model can comprehend objects, animals, places, and much more to bring your vision from text to an image. This allows it to convert a text query like: “A Balloon Dog, Under the Sea”.

Here’s how DALL·E 2 converts that complex text prompt into an image.

You might be wondering, “what is the difference between DALL·E 2 and DALL·E 3?”. According to OpenAi’s official website:

Even with the same prompt, DALL·E 3 delivers significant improvements over DALL·E 2.

— Dall-E-3

Currently, if you purchase ChatGPT Plus, you’ll get access to the Dall-E3 with the capacity to generate a large number of images from it. Dall E 2 Free is not available, but you can try out these 10 Free & Paid AI Art Generators to experience a text to image product.

How to use DALL·E 2

You can also directly purchase 115 Credits for $15 from Labs.OpenAi.Com and start using Dall E2 for image generation. If you do not wish to go ahead with the $20/Mo ChatGPT Plus Plan.

Step to use Dall E2:

1. Go to DALL·E 2’s web app and create your account
2. Click on “Buy Credits” and purchase 115 Credits for $15.
3. Enter your prompt on the home page and click “Generate“.
4. Within seconds Dall E2 will give you 4 AI-Generated images.
5. Click on the image that you like to either Download, Save to collection, Share to DALL·E 2’s public feed, or Edit.
6. In Dall-e make a variation from text commands to generate more such images. You can make variations of the same image by further prompting the AI with text commands

Also note that Dall E API is billed separately. The credits you purchase from Labs.OpenAi.Com are not applicable to DallE APIs.

Now’s lets dig deeper and answer a few questions for you:

What’s a DALL·E Credit?

• Think of your purchased Dall E 2 credits as a currency you spend to use Dall E 2 AI for each request: to generate image, edit image, or generate a variation.
• Your credits are utilised or spent only for prompts that return an image. So in case due to content policy warnings or errors, you don’t get an image, your credits won’t get deducted.

What are free credits?

• Earn adopters who signed up for Dall E before April 6, 2023 got free credits. But now they are no longer available.
• The catch was, the free credits expired for everyone after 1 month. Which means the Dall E Free Credits expired by 7th May 2023.
• All Free credits for Dall E replenish every monthly but that only happened till April 2023.

How do DALL·E credits work if I belong to a multi-person organization account?

• In this case, the free and paid credits are shared within each organization.
• The owners of an organization has the access to purchase more credits as they get utilized.

Is there a differences between free and paid credits for Dall E2?

• The expiry duration for Free Credits is one month.
• Expiry duration for paid credits if 12 months from the date of purchase.
• But both gets same rights to use the image output for commercial purchase, whether you generate images from free or paid credits.

Cost	Utilisation
~$0.13	Per Prompt
~$0.035	Per Image

Once you’ve purchased the credits, you’ll see the simple Dall E 2 screen with a prompt input box. Every prompt will generate 4 image variations. You can click on “Surprise Me” to let Dall E generate an image on its own to surprise you. Or you can try one of the following prompts in it:

1. “A futuristic cityscape at sunset, with flying cars and neon signs reflecting off glass skyscrapers.”
2. “An enchanted forest glade illuminated by bioluminescent plants and a clear, starry night sky.”
3. “A steampunk-inspired library with intricate brass machinery and shelves of ancient books, lit by flickering candlelight.”

Here’s the image output from Dall E 2 from the 1st prompt:

Now I can click on any one image, and I’ll be able to download or Save it to a collection. On Dall E 2, I also have the option to publish and Share it publicly with a link. I can Edit the image or create more Variations by further prompting the Ai.

Honestly, DALL·E 2 and DALL·E 2 are very easy to use — but we’ve a lot more to explore in it.

Build AI Image Generator App using DALL·E 2 from APIs

Startups, indie hackers, and companies that are looking to create some kind of AI Imager genereation capabilities as a a feature within their product can use APIs from OpenAi.

OpenAi encourages developers to build DALL·E directly into the apps to generate and edit images and art in real time. It acts as either a use-case (feature) or a wow-factor within the application. DALL·E 3 is the highest quality model and DALL·E 2 is a lower cost model optimised to keep the expenses low.

In the table below, you can see the various API pricing for DALL·E 2 and DALL·E 3:

Model	Quality	Resolution	Price
DALL·E 3	Standard	1024 X 1024	$0.040/image
	Standard	1024 X 1792, 1792 X 1024	$0.080/image
DALL·E 3	HD	1024 X 1024	$0.080/image
	HD	1024 X 1792, 1792 X 1024	$0.120/image
DALL·E 2	Standard	1024 X 1024	$0.020/image
		512 X 512	$0.018/image
		256 X 256	$0.016/image

How to write DALL·E 2 prompts for better image generation

To get the best output form Dall E 2, you’ll need to write good prompts. You can read Best prompts for AI Art Generators to explore some of those. But ultimately, you need to come up with original prompts for your own use cases to get the desired results. Here we’ll get deeper into a few factors that define a what is a better prompt for bette Ai image generation.

Apply the following tips for better results:

Be Specific with Prompts

If you just ask for “A Dog”, you will get weird results and the image might not have what you expect.

However, if you give a more specific prompt like “Realistic image of a dog jumping to catch a frisbee” you’ll get some interesting results:

Try descriptor and Styles

We’ve discovered that if we ask too much for a realistic image, you’ll not get great results in Dall E 2. However, we recommend that buying ChatGPT Plus and using Dall E 3 now gives you far better results.

You can have fun generating images to mimic certain painting styles, art work, sketches, or asking Dall E to generate images with certain vibes, era, or feelings. Almost all your crazy ideas you can think of to generate images on Dall E are worth trying.

Try a few variations of your text prompt or tweak it slightly to see how the ai model responds. Over time, with trial and error you’ll get a hang of it.

Don’t Complicate Your Prompts

If you complicate you text prompt with too many words, minute details, and make it elaborate, then DALL·E 2 ai will miss the focus. It may mess up the end result because a complicated text prompt. “An art piece of the Marvels and Avengers teaming up with Super Mario” is a fun ai prompt, but the resultant images might not be coherent.

Also, you need to understand that using brand names, copyrighted characters and symbols, or widely popular names might lead to a warning. Usually DALL E 2 is happy to generate an image inspired from the copyrighted material, so make sure you say that make an image inspired by “Your Reference”.

Even though your choice between the above images and the following images for a shorter and simpler prompt: “” might differ.

If you want more ideas, you can check out multiple free resources and prompts books on the Internet. The external prompts books are not a huge help because only you as a user or develop on Dall E 2 can know what are your requirements.

But you can try looking at this free DALL·E 2 prompt eBook. It breaks down multiple prompt formats and suggests how changing your prompt can impact the image output. No prompt bundle for ai is a replacement for trying out multiple things and finding out on your own that what works for your personal and work related use cases.

DALL·E 2 Advanced Features: Inpainting and Outpainting

Inpanting and Outpainting are only available in DALL-E2 via Labs.OpenAi.com. The inpainting and outpainting are not available in DALL·E 3 via ChatGPT.

But with DALL·E 2, you can additionally edit your AI Generated Image or editing your own uploaded using the following two techniques —

Inpainting

Inpainting allows you to erase some parts of your existing images or your AI Images. It can help you fill the gap in image, remove or replace specific parts, or alter select parts of it.

Outpainting

Outpainting feature allows you to do magic fill in the back. You can fill in the background, expand the background, or grow the image beyond the existing borders.

The best part is that you can also combine both of these feature and expand your image’s background and then inpaint in the new areas to magic fill the outputs.

How to Edit Images with DALL·E 2

To edit a photo or any image of your own save on your PC, here are the steps:

1. Go to DALL·E 2 home (labs.openai.com), Upload an image, and upload the image you want to edit.
2. Now you can crop the image into a square or you can skip this step.
3. Next, click “Generate Variations” to prompt DallE2 to use your own image. Or Edit Image to use advanced techniques.

Another option is to select any pre-generated image from Dall-E2 and click Edit.

How to inpaint with DALL·E 2

Once your image is open in the editor, choose the Eraser tool (the shortcut for it in the Keyboard is E).

Now, you need to paint over the part of the image you want to replace using the eraser tool, and use the Pop-up prompt bar to write your prompt (describe) what you want that part of the image to be like and along with information on what fills the selected area. Next, Click Generate.

The DALL·E 2 Eraser is a great feature to replace Photoshop or other eraser tools for open-ended editing requirements.

You will again get 4 options of the images to choose from. So probably one of the results will resemble with what you want in the output.

How to outpaint with DALL·E 2

Open an image in the Dall E 2 image editor by uploading from your PC or by selecting a previously generated image.

Select Add Generation Frame (the shortcut on the keyboard for it is F). Now you can position your cursor to expand the image in whichever direction you want.

Once you expand in that direction, write your prompt and click Generate. The outpainting has gone to another level in Stable Diffusion and Runway ML. The outpainting is a great feature in Dall E 2 ai, it lets your imagination run wild and helps you get wild results.

I used to Dall E 2 to outpaint the background of this Golf Ground.

In outpainting as well, you will get 4 options to select from. Click on the arrows to see those photos. If you like something, click on Accpet, if you don’t like you can Cancel it. You can also then add more frames to generate further or download the images you’ve generated.

The image editing in Dall-E2 is still in beta. The full image doesn’t get save. So you can download the additionally generated frames and create your own independent image. In case you miss downloading the edited images, you will need to recombine (rejoin) them in an app like Photoshop.

DALL·E 2 FAQs

Since Dall E 2 is widely popular and still getting adopted, there are naturally many questions around it. Here we are answering the frequently asked questions on Dall E Ai.