🪄A new Canva trick — Create images from text
🪄A new Canva trick — Create images from text
Here's how to make AI-generated visuals for free — Wonder Tools #136
It’s difficult to design creative images. But it’s easy to type a few words. That’s why the biggest buzz in tech this fall is about new services that turn text into visuals.
Using artificial intelligence, these almost-magical tools can create machine-generated images from any phrase you type in as a prompt. Want a bike-riding panda? A blue car flying across a rainy Manhattan skyline? Whatever you can think of, you can have in image form in seconds.
It started with a few niche tech projects like Dall-E 2, Midjourney and Stable Diffusion. Now it’s gone mainstream. Read on to see how these images look and to learn how to create your own with Canva and other new services.
Canva last week launched its own Text-to-Image generator. Here’s how to use it:
Visit TextToImage.app in a Web browser or search Canva for the “Text-to-Image” app.
Click to open the Text-to-Image app in a little drawer next to the project you’re working on.
Type in a phrase or use one of the suggested defaults to test it out. Wait a few seconds. You’re then presented with four image options.
Drag a generated image into your project.
Edit the image with Canva’s built-in photo editor to adjust it to your preference.
Download it or share a link to show off your new AI-assisted creation
Voila! You’ve officially entered into the robot age.
First, robots (algorithms powered by artificial intelligence, really) learned to write for you. I wrote about my surprising experience with that. Now, machines are generating images for us. Startups like Synthesia are generating AI video too.
Canva is now the biggest, but not the only service converting text to images. [Read my prior post about Canva’s recent relaunch with other new features.]
Alternatives to Canva for generating your own text-to-image visuals
To try out these services, I typed “a butterfly flying over a swimming pool in Times Square” and a few other prompts into new tools. Here are a few you can try.
NightCafe provides a few free credits so you can try generating images for free. Here’s the butterfly image it generated for me. And a visual of a “happy grandparent kissing a parrot.” You can customize your image by choosing from a variety of styles, and you can “evolve” or edit it after it’s created.
Stable Diffusion is one of the highest-quality services I tested. It generated these grandparent+parrot images for my prompt. You can try it free on this demo site or try the new DreamStudio beta for the full pro-level experience. Here’s their 100-second video demo on how to get started.
Midjourney describes itself as an independent research lab exploring new mediums of thought and expanding the imaginative powers of the human species. It’s run by a fully-distributed team of 11. See examples of what people have created in the community showcase. To learn more and join the beta, you can join a discussion group on Discord, which is akin to a Slack group.
Melobytes is a fascinating new service I stumbled upon that offers all sorts of AI creations, including Image-to-Song. It took the happy grandparent with parrot image I created and generated this bizarre 27-second sound clip. The service even has a feature to turn a subtitle file and turn it into a video. It also has a simple text-to-image service that yielded this tiny butterfly-prompt image. The service is not one I’d recommend for quality renditions, but its range of offerings signals what AI will eventually be capable of.
Craiyon is the first one I tried, originally called Dall-E Mini. In my tests the images were of lesser quality than the other services, as reflected in this rendition of my butterfly phrase. It’s free, though, and requires no log-in.
Some people are understandably concerned about this use of artificial intelligence. Witness the controversy over a digital art prize awarded to an AI-generated piece at the Colorado State Art Fair. What are your thoughts on how this AI can and should or shouldn’t be used? Leave a comment 👇 or reply to this post.
Sponsored Message
Running to Substack because Twitter is on the cusp of collapse?
Well, email isn’t your only option.
Winno lets you text your followers like they’re your friends. Think of it like Substack, with 90% open rates, a 40% CTR, and the ability to send short-form, long-form, videos, audio notes, and all kinds of content in between.
Sign up (for free) here to start creating your own paid texting community.
🙏 Thanks for Your Feedback on Wonder Tools
Many of you shared helpful input on the 3-minute reader survey I shared last week in my post about why I prefer Tally over Google Forms.
The survey is still open. Please fill it out if you haven’t yet. I’ve read every single response so far and I’ve begun responding to those who included your email address. An incentive to fill it out: see a new survey tool in action.
Your survey responses are sparking new plans:
You shared lots of great post ideas, from digital calendaring to “power combos” — apps that work well together. I’ve added many of these to my planning list.
Many of you expressed interest in short video demos and tutorials. I’ll experiment with these in the weeks and months ahead.
The Fresh Start in 2023 online course drew interest. The pilot course this January will be small. Fill out this form to reserve a spot.
I love Midjourney for actual art creation. I use it as the foundation of a multistep mixed media work now. But... I love that Canva might make basic IG and other image creation a step easier through AI.
I vote for the Notion AI text breakdown next week!