Canva's new image generator faces tough competition from OpenAI, google, and meta

Canva has launched its first image-generation model, Canva Craft, aiming to compete with tools from OpenAI, Google, and Meta. The model is available through Canva’s Dream Lab, accessible to free users with AI usage limits. Canva emphasized that Craft generates images based on design layouts, proportions, and text, focusing on legibility and accuracy. Unlike other models, it was trained on human-designed Canva templates rather than photographs. Business Insider tested Craft against ChatGPT, Muse, and Gemini, finding it struggled with simple tasks like designing a birthday card, producing awkward fonts and missing elements. While it performed reasonably well in more complex tasks, such as placing a pendant in a new setting, it failed to adhere to realistic physics. ChatGPT emerged as the top performer, creating a realistic and detailed image of a cat with accurate lighting and composition. Canva’s tool still has significant room for improvement, especially in graphic design, prompting users to continue manual adjustments. The release follows a previously published article announcing Canva’s new image model. This update highlights its performance compared to leading AI tools, showcasing both its strengths and limitations in image generation tasks.