Canva announced the launch of Craft, its inaugural foundation model for image generation, positioning the tool as a direct competitor to offerings from OpenAI, Google, Elon Musk’s xAI and Microsoft. The rollout on Friday marks the culmination of a six‑month development effort by Canva’s 160‑person Research team, according to co‑founder Cliff Obrecht.
Craft foundation model details
Obrecht told Forbes Australia that Craft is the first of many models the company plans to release, emphasizing that the research team built the system “completely from scratch” in roughly half a year. The model achieved a fourth‑place ranking on DesignArena, a benchmarking platform that relies on human judges to assess AI‑generated creative output. All three models ahead of Craft were developed by OpenAI, leaving Canva’s lab in second place overall.
Canva attributes its performance edge to the data set used for training. Rather than relying primarily on photographs, as many rivals do, Craft was trained on “millions of human‑made design templates” drawn from Canva’s extensive content library. The company also highlighted a two‑stage generation pipeline that separates composition planning from image rendering, a design it says yields more precise and consistent visual results.
Cost efficiency is a core part of the narrative. Canva’s AI tools have been used more than 38 billion times, and Obrecht noted that even a fraction of a cent saved per generation can translate into a substantial reduction in operating expenses at that scale. By owning the model, Canva claims greater control over both cost and speed, enabling it to scale the technology across its entire user community.
Market positioning and investor reaction
In April, Canva rolled out Canva AI 2.0, a two‑year overhaul of its platform described as the company’s largest product launch to date. The refreshed suite centers on conversational design, where users describe what they want and the system assigns appropriate models and tools to fulfill the request.
However, by July the firm warned investors that rising AI costs would delay some product releases and that projected revenue growth for the year had been revised downward. The market response was swift: Blackbird, Canva’s largest shareholder, cut its valuation of the company by 17 percent, bringing the figure to US$34.9 billion.
Obrecht framed the situation as an opportunity rather than a setback. He said the company hopes to “jiu‑jitsu the threat of AI into an opportunity,” aiming to monetize the model beyond its own platform. Canva intends to license Craft to other companies, allowing third‑party platforms and AI agents to generate images without directing users to Canva’s website or app. “It brings Canva to more people and opens up a whole new revenue opportunity for us,” Obrecht said.
Future roadmap and broader strategy
In a company blog post, Canva announced that it is already developing two upgraded versions of the model—Craft‑fast and Craft‑ultrafast—designed to improve efficiency over the original offering. The firm’s broader AI ambition is underscored by its recent acquisition of Leonardo.Ai, an Australian startup that previously held the distinction of producing the country’s first image‑generation model. Leonardo.Ai’s co‑founder JJ Fiasson now serves as Head of Generative AI at Canva.
Leadership of the research effort rests with Stef Corazza, who joined Canva last year after stints at Adobe and Roblox. Under his direction, the research team aims to expand the portfolio of generative tools that power both Canva’s internal products and external partnerships.
Overall, the introduction of Craft signals Canva’s intent to move from being a consumer of third‑party AI services to a producer of its own foundational models. By leveraging its massive repository of design templates and focusing on cost‑effective scaling, the company hopes to solidify its position in the rapidly evolving generative‑AI market while unlocking new streams of revenue.
Be First to Comment