Australian design app Canva is the latest creative platform to launch a text-to-image AI tool. The company began testing the feature in September and is now rolling it out to the app’s 100+ million users. The feature is an implementation of the open-source text-to-image model “Stable Diffusion” with additional security filters and a custom user interface to help guide Canva’s users to the results they want.
Canva offers a free application, as well as a paid version with additional features, that will give all its users the ability to generate 100 images per day with the tool.
Loading Canva’s text-to-image feature, you’ll be prompted to “describe the image you want to see” and show some sample prompts as inspiration (for example, “light watercolor of a koi fish in a pond”). You can then choose from a variety of styles (“photo”, “drawing”, “3D”, “painting”, “pattern” and “concept art”) and the tool will generate a grid of four images for you to choose from and add to your design canvas. There is also an option to report images that contain violence, nudity, hate speech and “biased and/or stereotypical” content.
A text-to-image tool generates four images of a panda riding a bicycle.
A screenshot of a text-to-image user interface that allows the user to select different styles: …
A screenshot of a user interface that gives you the option to report bad images: …
“We’re using this as a learning experience for our community,” says Cameron Adams, Canva’s co-founder and chief product officer.” We’re passionate about getting this technology in front of them because it’s an emerging field and the exact way it works and how customers will interact with it is still under development.”
The tool is already being used in a range of applications.” One of my favorites is that students use it to visualize their stories, so they’ll write a story in English class and use text-to-image to generate an image that matches that story. We’ve also seen it used for presentations, flyers and images on T-shirts that they can print through Canva.”
The feature is just the latest example of text-to-image AI tools reaching a growing audience. The launch of Stable Diffusion in particular has accelerated the use of these systems, as its open source implementation allows companies to integrate it into their own products for free. Text-to-image is rapidly becoming a staple of creative platforms, and just last month Microsoft launched its own text-to-image tool, Microsoft Designer (powered by OpenAI’s DALL-E system), as part of its Office suite.
The rise of these systems has also sparked some controversy, especially since they use copyrighted images as training data. Many artists have found that their work has been used to create these commercial products without their consent, even though the companies and researchers responsible say the use of this data is covered by provisions such as the U.S. Fair Use Doctrine.
When asked about these issues, Adams said, “I think there are legitimate questions about the extent to which AI products can be considered fair use, and that will vary around the world. We’re watching this issue closely, but it’s still all up in the air. We have a great relationship with our community of contributors and our users, and we’re working closely with them to figure out these copyright issues. We’re giving ownership of the images to our users, but we’re not claiming that the copyright on these images can be owned by those users.”
Bhautik Joshi, Adams and Canva’s “lead image and video expert,” emphasized that one of the key additions they made to the tool was the addition of filters to stop users from generating NSFW output – especially important if many of the users are school-age children.”
Joshi added that the company is “very aware that [the output] could be problematic, and that’s something we’re actively addressing.”