AI: Text to image comparison
DALL·E, Canva Text to Image, and Midjourney are three different AI-powered platforms that I have been using to date. In my last post, I shared the fact I am adamant AI will not replace me as an educator. With the speed at which AI is evolving – my head is actually spinning somewhat! Is anyone else feeling like this?
Nonetheless, the rapid pace at which AI is evolving has further confirmed that AI will not replace me because we (and our students) are entering a world where we inevitably need each other to navigate the changes that lie ahead. We need human interaction, collaboration, critical thinking and creativity more than ever.
In the previous post, I also mentioned the use of AI for text to image. So in this post I am going to share three of the main AI applications for text to image I have been exploring.
Canva: Text to image app
First up, the Canva text to image application. You can find this application in the ‘Apps’ tab on the left of the Canva workspace or by typing ‘Text to image’ in the search box. In short, this is a great little tool as long as you want simple common images that do not require close ups of peoples’ faces. In fact, I would avoid using it for any realistic photos; faces can appear quite distorted and quite frankly ‘creepy’. Canva for education has such great stock images, I find it quicker to find an image, use the one click background remover and add an element or text that I need to make the image specific for my needs. In conclusion, it has potential but this application has not made it into my daily list of apps to use.
Midjourney has both impressed me and driven me crazy! The images you can create are so varied in style and can be of very high quality. Images can be re-created from the first draft based on further prompts, so you can make it really specific to your needs. Below are three images I created on Midjourney for various purposes. The crazy parts – Midjourney uses the ‘Discord’ interface. To use Midjourney you must add text prompts to a public chat forum. You wait with multiple other users, watching multiple images pop up before your image gets created. When you finally find your image, you have a choice of four versions. When you request to upscale a particular version, you are again in the revolving waiting game with strangers; images of all sorts flash before your eyes. There are some workarounds to try to tackle this, but I do not deem this the most welcoming or timely way of creating your images. Nonetheless, if you were to put in the time, the images you can create really are quite spectacular. Unfortunately, Midjourney is no longer free and requires subscription on a monthly or yearly basis ($8-$10 a month).
Midjourney sample images:
Finally, the third text to image AI generator I have been exploring is DALL-E which is based on GPT. Similar to Midjourney, DALL-E can create some impressive images based on the prompts you input. It can generate images from a wide range of image types including realistic images, oil pastels and drawings. The interface is simple, clean and intuitive to use. DALL E easily lets you upload images for editing too. The main downsides are, similar to Canva, it is poor at creating people’s faces and it sometimes responds with “The server is currently overloaded, sorry about that.” DALL-E gives you 15 credits free a month which means you can download 15 images a month for free.
If you haven’t used AI text to image, maybe it’s time to try one out. Visuals, especially as we continue to venture further towards the need for online flexible learning, are becoming more and more important in how we present information to students. Hence, AI text to image has potential to help you create engaging high quality relevant images.
Cautionary note: policies are still to be developed and implemented with reference to the use of AI apps and data protection, so only input and expect output that you are happy to have potentially shared and accessible to others now or in the future.