Here's how Canva's new image generation tool compares to models by OpenAI, Google, and Meta (www.businessinsider.com)

Canva has made a significant move in the AI image generation arena with the launch of its new model, Canva Craft. This tool, integrated within Canva's Dream Lab and accessible even for free users (though limited by usage caps), aims to enhance design creation by planning layout, proportions, and text before generating images. Unlike many competitors, Canva's model is trained specifically on human-designed templates, such as invitations and presentations, instead of typical photograph datasets, which sets it apart in focusing on legibility and accuracy. Despite its innovative approach, initial tests alongside leading AI models from OpenAI, Google, and Meta reveal that Canva Craft has substantial room for improvement. In various creativity challenges, Canva struggled with prompt interpretations and accuracy, losing out to competitors like ChatGPT and Muse in key areas, such as realistic image generation and detail management. While it did handle some tasks well, like identifying elements creating atmospheric settings, it often produced visually implausible results. As the design platform continues to iterate on its model, its performance in the AI/ML community will be closely monitored.