Testing the DALL-E 2 generator – What is it capable of? 

Use AI to write high-performing marketing copy faster. 10x faster.

Test neuroflash's AI copywriter with the free plan.
We tried and tested the DALL-E 2 generator, the new AI technology by OpenAI, with our exclusive access. In this article, we will present our findings of what the text to image generator is truly capable of.
Share This Post

We got access to the DALL-E 2 generator a couple of weeks ago and have since tried and tested the technology. If you are not aware of what DALL-E 2 is or how it works, then check out our previous article on the DALL-E 2 OpenAI text to image generator, where we explain everything you need to know about the new technology. 

For this article, our goal was to gain more insight into how the image generator understands language and how different text prompts affect the AI generated images. We also asked ourselves how the DALL-E 2 generator could possibly be used in various fields like marketing and product design, including its strength and weaknesses. After exploring the potential of DALL-E 2, here is what we found: 

How the DALL-E 2 generator understands language - The duck toaster study

One of our first hypotheses was that the DALL-E 2 generator probably wouldn’t have any trouble generating images that you could also simply find on the internet by doing a quick search. For example, let’s say you want to generate pictures of cute ducks. Within just a couple of seconds, the DALL-E 2 generator offers you these AI generated images:

Source: OpenAI

Now, these are indeed some cute ducks, but you can easily reach the same results by typing “pictures of ducks” into a search engine. So, let’s get a bit more abstract and surrealist with our text prompt. Instead of generating pictures of ducks, we wanted to generate images of a duck toaster. When visualizing the results for the prompt “duck toaster,” we imagined that the DALL-E 2 generator would create images that merge the characteristics of both a duck and a toaster. However, by simply using the word “duck toaster” for our prompt, the results of the AI generated images were skewed.    

The artificial intelligence wasn’t able to comprehend what a duck toaster is supposed to be, so it utilized the words in the prompt that it recognized, like duck and toast, to generate a wide range of different images. Here, for example, the DALL-E 2 generator created an image of a duck with two pieces of toast:

Source: OpenAI

And here, in an awful twist of fate for the duck, the image generator created a picture of a duck that’s being toasted in the oven:

Source: OpenAI

It quickly became clear that the word “duck toaster” was not enough to generate the images that we had in mind. So in the next step, we got more specific with our text prompt and told the artificial intelligence to generate images of “a duck shaped like a toaster”. With this prompt, the DALL-E 2 generator got a bit closer to our expectations:

Source: OpenAI

However, the image generator still had trouble relating the attributes of a duck to the attributes of a toaster. This led to some misfires within the AI generated images and created pictures like a regular toaster with a duck-shaped piece of toast and a duck monstrosity that has grown some horrific machine-like appendages

Source: OpenAI
Source: OpenAI

Next, we switched up the words in the prompt to check if it will impact the results. Instead of “a duck shaped like a toaster,” we typed in “a toaster shaped like a duck” with the following results:

Source: OpenAI
Source: OpenAI

With the prompt “a toaster shaped like a duck,” the DALL-E 2 generator was able to create images that merged the attributes of a duck and a toaster more seamlessly, implying that the order of the words in the text prompt matters. Additionally, when adding a preferred artistic style to the prompt, like “a toaster shaped like a duck, Japanese color art” or “a toaster shaped like a duck, oil painting,” the DALL-E 2 generator was able to incorporate the styles without problem: 

Source: OpenAI
Source: OpenAI

In a final step, we decided to complicate the prompt even further by adding more details like actions. For example, we added “flying duck” or “holding a piece of toast” to the existing prompt and the DALL-E 2 generator started to include attributes like wings or pieces of toast in the pictures: 

Source: OpenAI
Source: OpenAI
Source: OpenAI
Source: OpenAI

Finally, we were even able to add emotions to the picture, like making the duck toaster happy or having it cry:

Source: OpenAI
Source: OpenAI

To sum up, the DALL-E 2 generator, just like artificial intelligence in general, doesn’t truly understand what it is doing. We as humans have to guide the AI into the desired direction. With the DALL-E 2 generator, we concluded that the more detailed the prompt, the better the results. It can help to switch around words to improve results, and you can add actions and emotions to make the images appear more dynamic. Finally, if you want to add a certain art style, it’s best to include it at the very end of the prompt and separate it from the rest of the sentence with a comma. At the end of the day, the fun part about DALL-E 2 is that you get to play around with it until you achieve the desired results. 

The future of the DALL-E 2 generator - Possible use cases

After we’ve tested the DALL-E 2 generator and witnessed what it is capable of creating, we started asking ourselves what possible uses the technology could have and how it could shape the future of various industries. With its capability to generate images from textual descriptions in different styles, including 3D models and photorealistic images, the DALL-E 2 generator could generate new ideas for a range of products or services:

DALL-E 2 generator creating associations in sales marketing:

Some would say that the most important aspect of sales marketing is understanding your target audience. Others might say that the most important thing is creating a strong brand. And still others might say that the most important thing is generating leads. However, one aspect that should definitely not be overlooked either is the power that visuals hold in the marketing world.

Images are important in sales marketing because they provide a visual representation of your product or service. They can help boost your sales by evoking emotions within your target audience and creating certain associations that people connect to your brand. A good marketing campaign thrives on unique visuals. Here are two hypothetical examples of how DALL-E 2 could be used for this purpose:

  • If the product you want to sell is candy, you could use DALL-E 2 to generate a visual ad of a brave explorer who travelled all the way to taste your delicious candy and holds it up like a priced possession for everyone to see:
Source: OpenAI
  • If you offer a copywriting service, then perhaps you want to associate your brand with intelligence, power, and pride. In this case, you could use the DALL-E 2 generator to create a visual ad of a lioness doing a writing assignment:  
Source: OpenAI

Fact is, with DALL-E 2 you can generate a variety of creative and unique visuals for your marketing campaigns that allow you to link your brand to certain associations and make it more attractive to potential users and customers. 

DALL-E 2 generator providing an edge in content marketing:

There are many important facets to content marketing, but some of the most important include creating quality content, tailoring your content to your target audience, and promoting your content through various channels. By creating quality content that is targeted at your desired audience, you can effectively market your business or product and reach a larger base of people. Additionally, promoting your content through social media, email marketing, and other channels can help to increase its reach and further engage potential customers or clients.

With the keyword for content marketing being “quality content,” it becomes clear that the content needs to be both unique and appealing. A large aspect of creating unique and appealing content are the visuals that people see even before they’ve read through the content itself.

Let’s say, for example, that you have a blog where you want to market content about AI technology to your target audience. If you plan on including images that showcase the relationship between humans and AI technology, then your options are limited. You either have to pay for a digital artist or stock photography, or find copyright free images that a lot of other articles from competitors have already used as well. However, with the DALL-E 2 generator, you can get creative and generate a vast number of distinctive and one of a kind pictures that fit your content in. Here are some examples that only took a simple text prompt and a couple of seconds to generate:

Source: OpenAI
Source: OpenAI
Source: OpenAI
Source: OpenAI

This way, you’ll never run out of creative and original pictures that you can add to your content to increase its quality, plus it will save you a lot of time. 

DALL-E 2 generator revolutionizing product design:

In product design, it is important to consider the customer experience and how the product will be used. It is also important to consider the manufacturability of the product and how it will be assembled. However, on top of all of this, it matters what the product will look like as well because the design is the very first aspect of the product that customers encounter and judge.

In order to sketch a creative product design and visualize it for production, design teams require the work of talented digital artists. Such artists require resources and time to do their work and they’re only capable of creating a limited amount of possible design choices at a time. The DALL-E 2 generator could revolutionize this process by allowing design teams to test out various design choices by simply generating them with the DALL-E 2 technology. 

For example, the market for sports shoes is incredibly profitable but also extremely competitive. To win customers over, the shoes don’t just need to be high quality, but they also need to have an interesting design. With the DALL-E 2 generator one could easily type in a text prompt of a design they have in mind and creat realistic pictures. Here are some examples for the prompt “basketball shoes in the style of maximalism, product photo”:

Source: OpenAI
Source: OpenAI

However, the product design can also be more specific, like adding a rainbow pattern to a pair of sneakers:

Source: OpenAI
Source: OpenAI

Fact is, the DALL-E 2 generator can create pictures of products to visualize possible designs. These pictures can also be realistic enough to be used as product photos on the website of the business.  

The limits of the DALL-E 2 generator 

With all the potential advantages that the DALL-E 2 generator has to offer, we finally asked ourselves where the image generator had its weaknesses. Sure enough, one of the first limits that we encountered was when we tried to generate realistic pictures of people, like in the context of a stock photo. Since the DALL-E 2 generator creates images from multiple existing sources, the faces of the people will oftentimes turn out distorted or plastic-looking in the AI generated images. 

Source: OpenAI

Another aspect that became apparent really fast was that the DALL-E 2 generator hasn’t learned how to spell properly yet. Instead of the words “copywriting” and “coffee,” the AI comes up with a jumble of words: 

Source: OpenAI

This also makes it impossible to create images with infographics or brand names. For example, when trying to generate an inforaphic about vegetables and healthy eating, the design itself looks good, but the text is unreadable. The same goes for trying to include a brand name. When trying to generate a visual ad for the Credit Suisse bank, the design looks good again, but the name of the brand is incoherent:

Source: OpenAI
Source: OpenAI

In the end, we’ve found that you have to be careful when using the DALL-E 2 generator, as the guidelines and rules are very strict. If you type in a word that is a content violation, even in a different context, then OpenAI will mark it as a strike. For example, when we tried to generate the Mario character “Bullet-Bill,” the image generator marked the word “bullet” as a violation of the guidelines. If too many of these violations occur, the account will be automatically banned. 

The future of the DALL-E 2 generator

So far the DALL-E 2 generator only offers limited access to a number of users, as it is still in the stage of a closed beta version. However, there are also free versions like DALL-E Mini, which is an attempt to recreate the success of DALL-E 2 with an open-source model of the technology that is accessible to everyone. The quality of DALL-E Mini may not be as good as the quality of DALL-E 2, but it is still in training and improves steadily. 

Interestingly, competitors like Imagen, Google’s text-to-image diffusion model, challenge OpenAI’s image generator by promising an “unprecedented degree of photorealism and a deep level of language understanding”. The competition between different brands of text to image generators might promise a faster pace in the development of such technology and might prompt companies to switch to open access software earlier.

All in all, it can be said that the DALL-E 2 generator has a promising future ahead of itself, as OpenAI is planning to continue its improvement and development of the DALL-E 2 technology. The DALL-E 2 generator might soon be able to create even more realistic and accurate pictures in the near future, though it can be said that it probably won’t completely replace digital artists, since they can reference specific technical terms and art historical words which gives their results more specificity and control than someone just trying to make something “cool” with the DALL-E 2 generator.

Finally, an interesting future development might be the fusion of text to image generators like DALL-E 2 and other AI technologies like GPT-3 text generators, which are already widely available. GPT-3 text generators can do a lot of things. They can help you with your writing, by providing suggestions and ideas. They can also help you with your research, by finding and organizing information for you. By fusing these text generators with the DALL-E 2 technology, AI would be capable of creating entire texts together with unique AI generated images. You can check out what a GPT-3 text generator is already capable of by itself with neuroflash. You can use our text generator to create over 80 different text types in different languages and up to 2.000 words for free each month, no subscription needed. 

Generate unique AI images with neuroflash

With the ability to generate images from text, the potential of artificial intelligence as a resource becomes clear. This is great progress achieved thanks to modern technology. That’s why neuroflash now combines the #1 German-language text generator with a new feature, text to image generation. This makes neuroflash the first company in the DACH region to offer its customers the opportunity to try AI image generation for themselves for free. 

How can you make money with neuroflash’s AI generated images and use it for your business?

  • Low-content books
  • Covers for books, songs, comics, e-books, …
  • Illustrations for bedtime stories, books, comics…
  • Print on demand images or post cards
  • Easy stock images for blogs (e.g. food blogs)
  • NFTs
  • Presentations & Slide decks
  • Images for social media posts, Newsletters
  • Inspirations for landingpage designs, product designs

Tip: Combine neuroflash with Photoshop or other programs:


  • Use Photoshop and enlarge our images from 72 dpi to 300 dpi with Preserve Details 2.0.
  • In the latest version of Photoshop (Beta) there is a “Photo Recovery” feature under “Neural Filters” that usually improves the appearance of the eyes and other odd facial features.
  • Then do some basic curve and color corrections and brighten the eyes a bit.

Use AI to write high-performing marketing copy faster. 10x faster.

Test neuroflash's AI copywriter with the free plan.

More To Explore

Content Creation

New Professional Plan – ChatGPT Pricing

What is the new ChatGPT price and what does it offer? What’s next for the popular chatbot? In this article you will learn what changes with the new price and how you can react to these changes.

Content Creation

ChatGPT Update: The new ChatGPT Professional

ChatGPT Professional is the new feature of the well-known and popular ChatGPT. In this article you will learn what to expect from ChatGPT Professional and how to get access to the new plan.

Test your magic pen in action:

Step 1: Choose a text type
Step 2: Enter briefing
?

No more writer's block

neurolfash-aiwriter

Use our AI to generate new content within seconds for free.