The design unicorn Canva has rolled out its brand new AI image generator on its website.
This feature began testing on Canva’s online image editor back in September, but it’s now available to all users –free and paid– across the Canva platform including web and installable apps.
This makes Canva the latest in officially incorporating text-to-image generation technology in its offer, opening up creative possibilities for its non-designer customers, and keeping its service up to date with the times. Try Canva’s AI image generator here!
Just like we reported back in September when the beta test started, this new AI image generation feature from Canva is built on the base of Stable Diffusion’s open-source technology with additional tweaks to tailor it to Canva’s concept and interface.
You can access it from its own page on Canva’s site, or from the “Text to Image” button on the left-side panel in Canva’s editor.
The tool gives you a text field where you have to simply enter a text prompt describing the image you want to see – you will find examples of text prompts as a guideline for how to construct them, “a panda riding a bike through a city with depth of field”, for example.
Plus, you also get a few options for a visual style: photo, drawing, 3D, painting, pattern, and concept art are the ones available at this time. You can also use “surprise me” to let the software choose a style on its own.
Once you enter your descriptive sentence and hit the “generate image” button, it’ll take a few seconds until you see four AI-generated images following your specs. You can then choose the one you like to add to your design, and further fine-tune it using Canva’s photo editing features such as modifying brightness or contrast, applying filters, adjusting focus, and more.
The beta test used to give 24 image generations per user per day, but now they’ll take it to 100 per day.
The feature is available for free users as well as subscribers to Canva Pro –the premium service of Canva.
Among the adjustments that the platform made to make this feature (based on Stable Diffusion) more aligned with its users are:
- Filters that prevent users from creating inappropriate or “NSFW” content. Something particularly relevant since children use Canva.
- A report button that lets you inform the Canva team when an image contains violence, nudity, hate speech, or biased and/or stereotypical content.
Plus, the feature has a very simple user interface design that fits within Canva’s user-friendly philosophy.
As we have been telling you, one of the core issues around AI images is that of copyright and legitimacy. Stable Diffusion just like most other AI image generation developers out there has and continues to train its algorithms with billions of images from the web, many of which are copyrighted.
That brings the question of, are the images generated by this tool copyrightable. According to Canva’s FAQ, they cannot guarantee it. The website specifies that “Our platform has no licensing or copyright claim over any work created using Text to Image”, and advises users always to disclose when they publish a design containing AI-generated images.
In short, they give you ownership of the content you generate with the new feature, but they don’t guarantee that you can copyright it, and they take no legal responsibility for the content at all.
However, Canva’s co-founder and CPO Cameron Adams told The Verge that they’re aware of all the issues surrounding the copyright of AI images and that they’re working closely with their contributors to figure out answers to the copyright questions.
For the time being, the tool is still being tested, and where it’ll go from here depends a lot on what Canva users do with it.
Have you tried Canva’s Text to Image feature yet? What do you think?