AI tools for text to image generation

Introduction

I learned about AI tools for generating images from a text prompt, only a few days ago so I had to try it out. I didn’t realize that these tools had progressed so much in the past few months. It turns out that similar technology can also be used for other things like creating videos from text or speech. Also things like writing content for web sites or blog posts. Maybe I will try that for a future blog post. In the mean time, I have been using the tools below to create various images which I have converted to NFT collections, banners, avatars etc. In fact the image that I attached to this blog post was generated with one of the tools. You have to try it out to get an idea about it works, and how easy and fast it is. Its just going to improve further and the possibilities are endless.

Here is a recent article which talks about how to get started and some of the alternatives:
https://creatoreconomy.so/p/how-creators-can-get-started-with
Here is an article about some of the ones out there:
https://beebom.com/best-ai-text-to-image-art-generators/
Here are some of the images that I converted to NFT’s :
https://fblauer.nftz.me/
https://solcial.io/users/fblauer
https://opensea.io/collection/rockstarz
Theses are all of different chains. (I hope you can see them without connecting a wallet) That's the subject of another article.

Tools to create images from prompts

You can install stable diffusion locally since it opensource, but I wouldn't recommend it unless you are an advanced user. Also it requires a a high powered GPU which most people don't have. So you are better off using one of the web based services below:

Imagen — This is the google research project which seems to be very advanced. But its doesn’t seem to be available to the public yet. Interesting to read about nonetheless.

Diffusion/Dreamstudio — This the opensource model from Stable Diffusion is opensource, and has good API's. So, this is the one that most web based services are based on and I used them to generate most of my images. You get a certain number of credits which you can use for trying it out, and there are various payment plans. I have been using it thru the Playground service below, which gies you 1,000 free image generations per day for free.

Dall-E — This also a really good model which comes from OpenAI labs. It will give different results so I have been testing different models. Also quite sophisticated images are generated. Many times it surprises you and gives you something you didn’t expect. This one is not free.

MidJourney — This one has a beta version which you can sign up for which gives a certain number of credits before you have to start paying a monthly fee. It all works inside of discord which has certain advantages and disadvantages. If you are already using discord, it is pretty easy to get up and running. Personally, I am not a big fan of discord and I don’t think it is the best approach for this type of app, but it is very popular and gives good results. You can try it out if you want to see how it compares. There is a monthly fee after the trial version expires.

Playground — AI — I found this one was the easiest one to get started with and it will work with multiple models. (SD and also Dall-e) You can just copy the prompts from the playgroundai page, and copy them into the create screen. You can modify the prompts or settings, and re-generate the images to see what effect it has. You can use PromptMania to see all the valid key words that can be used in the prompts.

Mage - Here is another one similar to Playgroundai which produces high quality results, and is free to use for now. You can explore images that other people have created, and re-use the prompts. You can click "View options" on the first page to try different settings, as well as prompts and see what kind of results it produces. You can also follow other people. So, if you follow me you can see lots of experimenting that I have done. You can also chose to make your images private, if you plan to sell them as NFT's.

If you are using SD 2.1 its a good idea to use negative prompts so that it won't give you deformed images for example. I will give you some example of negative prompts in future blog posts, where I get into more detail.

All of the various tools have lots of documentation, and social communities which you can connect with. I found the best way to learn was to read a little, or watch a youtube video, and then spend some time experimenting. Once you get the hang of it, its addictive and I have spent a fair bit of time with it in the last few days, trying things out and looking for use cases. See some of the links above for examples of what I have produced.

Conclusion

So, Everyone is asking the question if graphic artists are going to lose their jobs. Like most technology it is good at certain things and weaker at others. For example, if you have something exact in mind, you may not be able to reproduce it exactly the way you want. But if you are flexible, and just looking for new ideas, it is a useful tool. It also depends on what you are using it for and your skill level, both technical and creative. Some use cases will work better than others. To answer the question, I really don’t think that graphic artists are in any danger of being replaced, any more than a drummer will be replaced by a drum machine, or lawyers will be replaced by smart contracts. But this could definitely be a valuable tool in the right hands and it might even end up generating more work. As far as I am concerned you still need an artist to come up with something really good. Try it out, and let me know what you think.