Canva Craft AI Image Generator Launches With 2K Resolution

Canva has launched Craft, its first in-house AI image generation model, featuring up to 2K resolution, improved text legibility, smarter image composition, and targeted editing. Available through Dream Lab, Craft can also turn AI-generated images into editable designs, with planned API access and integrations with AI assistants such as ChatGPT and Gemini.

---

Canva recently rolled out Canva Craft, which marks the first AI image generation model built by the company in-house. The new AI image generation model represents a fresh way of making visuals using artificial intelligence. Canva Craft allows generating more polished and visually pleasing images that have better composition, more readable text, and more control over each individual element. The new model supports resolutions up to 2K, more than 30 aspect ratios, and selective editing capabilities, helping users get the most out of AI-generated visuals.

While traditional AI image generators just interpret the text prompt provided by the user and generate the image accordingly, Canva Craft involves an extra step of planning. As per the statement made by the team from Canva, the image generation process starts with the model coming up with a plan that involves taking into account spatial relationships between objects, the size of objects, visual hierarchy, and location of text.

Text placement is especially vital for people designing such things as posters, advertisements, graphics for social media, presentations, etc. where typography plays a crucial role. As far as AI images are concerned, they have been known to struggle with text readability and correct arrangement of elements. This tool called Canva Craft is created specifically to overcome this problem by taking into consideration the structure of the picture that should be generated.

One of the most interesting features about Canva Craft is targeted editing. Instead of regenerating the whole picture each time you want to make some changes, it allows you to edit individual elements while leaving the rest of the picture as it is. In this way, you will be able to make some corrections and adjustments to the picture that you are working on. It supports around 30 aspect ratios.

Canva Craft Uses Human-Created Designs for Training

One of the most distinguishing features of Canva Craft from other image generation models is the way that it was trained. The difference here is that while most image generation models are trained using photographs, Canva claims that the images in their model were learned from examples of human-made design templates taken from their huge database of design materials. This includes different designs of posters, social media posts, presentations, infographics, menus, and layouts featuring intensive typography.

This training technique is aimed at helping the model to understand how professional designs are organized. Training on design templates helps the model to learn how images, text, shapes, and other graphic elements work together. This training, according to Canva, improves object accuracy, text legibility, and composition, which are very important when designing images that should also carry some informative value.

With an emphasis on training for design, Canva Craft can prove to be especially beneficial for those people who constantly make marketing content and graphics for branding. For instance, one could make a promotional poster, another company could make a social media advertisement, while a third one could make an infographic and not worry about placing every single element of their creation manually. The final result would still depend on the prompt and the level of complexity of the design, but that is exactly what this model is built for.

Turn AI-Generated Images Into Editable Designs

The feature will also work in conjunction with the image-to-design functionality of Canva Magic. Through the use of the technology, the generated image can be turned into an editable and layered design. As such, it is possible to go past the use of a single image in the process and engage in a more flexible design creation process.

The feature will be beneficial to individuals who wish to work further on generated visual materials. The users will not have to begin a new process in the design editor but turn an image into a design and edit its elements individually. As such, the feature will complement Canva's approach to combining the generation process with the usage of its editors in order to create a final design.

The feature will make it easier for the content creators and small companies to work with AI-generated material. The users can generate the image, edit its certain parts and continue working on the design using the tools provided by Canva.

Why Canva Is Building Its Own AI Image Model

Before launching Canva Craft, the company used the external AI models for images generation. Building the internal model allows the company to get more control over the working of the technology as well as over the integration process.

The internal model can help Canva to deal with generation costs, to improve performance and to customize the process of images generation in order to meet better the needs of its customers. The internal model provides more flexibility for further improvements compared to reliance on the third party. With the growing popularity of AI-powered design solutions, Canva can benefit from having the image generation technology in-house.

At the same time, the building and operation of such model involves constant costs for training, infrastructure and optimization. The long-term success of Canva in using the internal image generation technology will depend on its performance and usefulness for users.

Canva Craft Is Coming to the API and AI Assistants

Craft is also going to be available on the Canva API, which may give developers an opportunity to integrate the image generation abilities of this model into their own apps and services, thus expanding its usage range beyond the Canva website and making it possible to utilize the design-focused generation technology from other software.

Moreover, the company has announced that it wants to implement integrations with AI assistants, including ChatGPT and Gemini, using the Model Context Protocol, or MCP. MCP is a means of enabling compatible AI applications to integrate with third-party tools and services. In case these integrations will be implemented as planned, users will be able to leverage Canva's image generation capabilities from assistant-powered workflows rather than using only the Canva interface.

This type of integration would come in handy for those interested in combining conversational AI and the process of creating visual content. For example, a user can work out a campaign idea using an AI assistant and then use the Canva features connected to the assistant in order to create the necessary graphics.

Canva Craftfast and Craft-ultrafast Are Also in Development

In addition to Canva Craft, the firm is working on two other versions known as Craftfast and Craft-ultrafast. These versions are expected to offer greater efficiency in terms of costs in Canva's image generation technology.

Such efficient versions would enable Canva to cover more types of creative activities while controlling the computational costs related to AI image generation technology. These versions can also offer more leeway to Canva in terms of achieving a balance between quality, efficiency, and cost. Nonetheless, specifications of Craftfast and Craft-ultrafast versions have yet to be disclosed.

At this point, the Canva Craft is a significant milestone on the path to creating their own artificial intelligence technology for visual design. By focusing on the aspects such as layout planning, typography, targeted editing, and editable designs, the tool takes an approach that stands out against other similar tools designed primarily to create images. In addition, by including plans to provide access to its API and integrate it with ChatGPT and Gemini, the company is preparing its image generating technology for application outside the design platform.