The Top 3 AI Text-to-Image Tools

How to Choose the Best AI Text-to-Image Tool for Your Experience and Budget

AI image-generation or text-to-image tools allow users to enter a text prompt as they would with a chatbot and the program will produce an image or multiple images based on the prompt. As with prompts for chatbots, the more well-structured and detailed your prompt the more likely you are to achieve the desired result.

These tools not only allow individuals and businesses to generate unique and custom images but also open up new possibilities for data visualization and even art creation. Here are just a few ideas of how these tools will be used:

• Allow individuals and companies to create royalty-free images for use in marketing materials, for their blogs, and for their social media pages.

• Allow users to create instant mock-ups for things like new product lines, construction or home projects, and even fashion design.

• Data visualization will also be another use case. Imagine you could provide a text-to-image generator with a table of data and ask it to create a unique visualization of that data that anyone could easily understand. That’s powerful!

• People will also use these tools to create art. As we’ve already started to see, AI-generated art can be versatile and convincing as real art, and who is to say it isn't? One day we just might have an artist who becomes famous specifically for their AI-generated art.

In this post, we will cover some of the key players in this space, Midjourney, DALL-E, and Stable Diffusion, and discuss the benefits and drawbacks of each and how to use them most effectively.

The Key Players in Text-to-Image AI Tools

Midjourney

Midjourney is arguably the most well know text-to-image generator. Its most differentiating quirk is that users interact with it through the popular social platform Discord. While this makes it a bit less user-friendly than its competitors it does add more of a community component to the generator.

To begin using Midjourney, you will need to sign up for a Discord account and join the Midjourney channel. Midjourney provides a helpful quick-start guide to help you complete your setup here.

Midjourney previously offered new users 25 free queries before prompting them to switch to a paid plan. However, at the time of this writing, the free access has been halted due to excessive demand. As far as paid plans go, users can choose between the Basic, Standard, and Pro plans, which allow larger user queues, more GPU time, and the ability to work via direct messages with the Discord bot to give you a private space to work.

To use Midjourney you will enter your prompt as a command in Discord using “/imagine”. Within a minute you will receive four images for your prompt.

Once these images are generated you will see nine buttons below the images “U1, U2, U3, U4, V1, V2, V3, V4,” and a re-roll button.

The letter “U” stands for upscale and the number attached to it identifies the photo that you want to upscale. “V” stands for variation and, once again, the number identifies the photo that you want to generate variations from. Re-roll can be used if you want to use the same prompt but have four new images generated.

Upscaling will generate a larger 1024 x 1024 copy of the selected image and will add more details. Overall it should provide a more finished look to your image but it is important to note that some details may change.

After upscaling you will see a few new options pop up. The “Make Variations” option will create four variations of the image you have upscaled. The “Beta/Light Upscale Redo” will utilize a different upscaler model to upscale your image. Light Upscale Redo will smooth out the upscaled image to give your final image a more polished look. Beta Upscale Redo will allow you to create a larger image size but will likely produce a noticeably different look and feel to your image.

Variations can be useful if you like one of the images that Midjourney produces and want to see additional, similar options to that image.

If you don’t like any of the images, consider using the re-roll feature or entering a new prompt altogether that better suits your desired output.

Conclusion

While it might not be as easy to initially access as its competitors, the setup for Midjourney is actually quite simple and the chatbot is easy to use. Once you have it set up you will find that its large and active user base provides a robust community, helpful resources, and reliable support, making it a great choice for your go-to text-to-image generator.

DALL-E 2

DALL-E 2 is the second iteration of OpenAI’s image and art generator DALL-E. The most notable difference between the two is that DALL-E 2 can create more high-quality images in less time than its predecessor. It also allows users to request edits to the images produced by the program in what OpenAI calls “in-painting”, where certain aspects of an image can be edited or replaced.

DALL-E 2 allows users to use an image as an input and get variations on that image, whether it be alternate angles or different styles of the same subject.

The program was trained using images and their text descriptions and learned to link images together allowing it to process odd prompts like a kangaroo hitting a golf ball.

If a user requests DALL-E to create an image using a text description of something it has not been trained on, the system will try and create its best guess of what that image would look like, however, it can create faulty images. While the system was trained on a robust data set, it may have knowledge gaps that will cause users to receive inaccurate images. For this reason, it is important for users to review the images they generate for accuracy before using them for any official purposes.

DALL-E also has a neat feature called “outpainting” where it can take an image input and create additional context around that image. For example, you could take the famous painting Nighthawks by Edward Hopper and have DALL-E create additional cityscape around the diner, which is the subject of the original painting.

DALL-E uses a credit system for image generation. Users who signed up for DALL-E prior to April 6th, 2023 get free credits that replenish monthly. Any users who sign up after that time will have to purchase credits to use the program. Credits can be purchased via the DALL-E website in increments of $15, beginning with 115 credits for $15. Each credit represents a single request of the program.

Conclusion

While DALL-E does not offer a free option to new users, the program is still very affordable at about $0.13 per credit/request. It has an easy to user interface and produces quality images with some neat features to get iterations and edits on its outputs.

Stable Diffusion

Stable Diffusion is another one of the most popular text-to-image generators. The platform prides itself on creating high-quality images in little time while maintaining user privacy by not requiring or using your personal information, or storing your images or text prompts.

The platform was trained on a data subset provided by a German charity (LAION), which did a general crawl of the internet, providing Stable Diffusion with a broad dataset to work from.

The platform does not require you to sign up and is currently free to use, without any features behind a paywall. Anyone can use it to its full potential, as long as they know how to create solid prompts. So let’s discuss how it will help you do that.

Stable Diffusion offers two specifically interesting features, which are the Prompt Database and the Prompt Generator.

The Prompt Database allows users to search topics and see images that have been created around that topic and the prompts that were used to generate them. This allows users to find inspiration for their own prompts and to see what descriptions work best to reach certain outcomes. Another way to improve your prompts is through the Prompt Generator feature.

The Prompt Generator is aimed at helping users by taking their existing prompt and tailoring it to be more precise and usable by the program. The idea is that it will help users create a prompt that is more likely to get them their desired image from Stable Diffusion. After all, like all of the platforms discussed here, what you put in is what you get out, and the better your prompt, the better your image will be.

Conclusion

Stable Diffusion is without a doubt one of the most accessible text-to-image generators available. Its ease of use, speed, Prompt Database and Prompt Generator tools make it one of the best platforms for new users to experiment with. Not to mention they can use it for free!

Final Thoughts

Choosing which platform to use depends largely on what you are trying to accomplish and what your budget is. If you are new to text-to-image generation and just want to experiment, we would recommend Stable Diffusion as it is easily accessible and free to use. If you want to use text-to-image for more professional purposes DALL-E is our top choice due to its easy-to-use features and high-quality outputs.

Overall, aside from the platform you use, the most important part of getting good image outputs is how you prompt the program. Looking at the work of others and practicing variations of your own prompts can be a great way to learn. There are great communities around the three platforms we covered so you can be sure to find great resources on Discord, Reddit, and Twitter.

Whether you use text-to-image tools for fun or for professional purposes, they can be truly powerful and will become more commonplace in the coming years. Starting to experiment with them now will help you learn how to use these tools effectively, save time and effort, and unlock new creative potential. No matter your goal, get out there and see which platform you like best!