Canva integrates a new graphic creation tool powered by AI. Called Text to image, it generates visual content through a simple textual description. A somewhat controversial technology…
Image generation artificial intelligence (AI) is about to revolutionize the world of graphic design. Now, Internet users can create photorealistic scenes, real works of art or incredible 3D renderings by writing a simple sentence thanks to software like DALL-E. It’s the turn of the Canva design platform, which allows you to create fairly elaborate visuals and flyers in a few clicks (see our practical sheet), to launch a new “Text to image” function, which is based on this principle. It started testing this new tool in September and has just rolled it out to its users around the world. Now, just enter a simple description in a panel to instantly create multiple designs, which can then be incorporated into invitations, cards, social media posts or flyers, without the user having to know in the matter. A tool that promises to be very practical but needs to be improved, not to mention that it raises some issues.
Text to image: images created from a simple description
To develop this new function, Canva relies on the open-source software Stable Diffusion. It is based on deep learning, a technology that mimics the human neural system through algorithms. By analyzing and cross-checking models of millions of photos, he manages to understand which word belongs to which image. It was then equipped with additional filters – to avoid generating racist, sexual or violent images for example – because the basic ones are quite easy to circumvent, and its interface has been personalized. The new tool is available on both paid and free versions of Canva, with a limit of 100 images per day for each user.
To use the tool, just “describe the image you want to see”. Canva gives a few examples for those lacking inspiration, like “a light watercolor painting of koi fish in a pond” or “a panda riding a bicycle through a city with depth of field.” Then, you have to choose the desired rendering style – drawing, painting, pattern, 3D, concept art and photo – and the tool is responsible for generating four unique images, which can be integrated into the creation. For now, the tool is nowhere near as powerful as those found on the Internet. Indeed, you have to enter a fairly simple description and there are currently a lot of failures. And when it works, the rendering is quite… strange and frustrating. “We are treating this as a learning experience for our community”said Cameron Adams, co-founder and chief product officer of Canva, in an interview with The Verge. “We’re keen to introduce them to this technology because it’s an emerging area, and exactly how it works and how customers will interact with it is still under development.”
The Text to image function is directly integrated into the Canva interface, along with the Photos, Text, Templates, etc. tools.
► Go to the Canva site with your web browser or open the Canva app on your mobile.
► Once on the Canva home screen, create a new document by clicking on the purple button Create a design then selecting the desired format. You can also resume one of your creations.
► In the toolbar on the left, click on the last icon, Apps.
► A panel opens. Type Text to image in the search bar and click on the purple icon of the same name.
► Canva briefly explains the function of the tool. Click the purple button Utilize a menu base.
► A new panel opens. Start by entering the description of the image you want to obtain in the field provided for this purpose (colors, objects, environment, style, people, etc.)
► Indicate the type of illustration you want by selecting the corresponding duck (drawing, photo, painting, 3D…).
► Click on the purple button Generate an image.
► Wait while the AI is working.
► The software offers you four images. If you are not satisfied with the result, click on Restart and change your description. Otherwise, click on the proposed image so that it appears on your document. And There you go !
AI image generation: revolutionary but controversial tools
AI image-generating tools are becoming more and more accessible to the general public. The launch of Stable Diffusion has made it easier for companies to access this technology since it is available in open-source. They can therefore integrate it easily and free of charge into their software and services. Canva is not the only platform to try it since Microsoft has announced the arrival of a new graphic creation tool working thanks to the AI of DALL-E called Microsoft Designer. It will be integrated into the Microsoft 365 suite when it is released. However, the rise of such systems is a source of much controversy because the training of AI is based on the use of millions of images, some of which are protected by copyright. The works of some creators have been used without their consent and are used to “feed” the AI, which will then create products for commercial purposes. Moreover, these tools generate moral and ethical questions. Who is the author of the created image, the AI or the user? Can an AI be recognized as an author? And, more generally, can an AI be considered the creator of a work?
Asked about it by The Verge, Cameron Adams said: “I think there are legitimate questions about the extent to which AI productions can be considered fair use, and that is going to differ around the world. We are monitoring the situation closely, but everything is still very uncertain. We have a great relationship with our community of contributors and our users, and we work closely with them to resolve these copyright issues. We give ownership of the images to users, but we do not claim that they can be protected by the copyright of these users.”