What Canva announced
Canva, a design platform, has introduced Canva Craft, its first image-generation model developed in-house. According to a report by The Decoder, the company claims the model can compete with similar products from Google, OpenAI, ByteDance and Meta. The launch marks a shift: until now Canva integrated third-party tools, whereas it is now bringing visual generation into its own infrastructure.
The company says the model was trained from scratch by its internal lab, Canva Research, using mainly millions of design templates instead of photographs. It is a choice that reflects the nature of the platform, which is oriented toward graphic materials and compositions rather than pure photography. Researchers from Leonardo.ai, which Canva acquired in 2024, also contributed to the project, according to the source's account.
How it works and where it is available
Craft is available to all users through Dream Lab and the public API. Canva plans an MCP integration for chatbots such as ChatGPT and Gemini later, but for now this is an announced plan and not a feature that is already active. Two lower-cost variants, called Craft-fast and Craft-ultrafast, are also expected soon.
The claims and the limits to be verified
Canva says that, in the DesignArena platform ranking, Craft placed fourth in the overall evaluation and first in portraits and landscapes. However, these are results reported by the company itself and not independently verified in the available material. Likewise, the comparison with models from Google, OpenAI, ByteDance and Meta is a statement by Canva, not an outcome certified by third parties.
The model's in-house control is presented as an advantage in quality and speed, as well as in costs. These are plausible claims but not quantified in the source: no prices, generation times or comparison metrics are provided. The real performance, the quality of multilingual text and the actual competitiveness against the cited competitors therefore remain to be verified.
Why it matters
Canva's entry among producers of its own models expands the number of players in the sector and shifts the company from the role of integrator to that of developer. For users of the platform, this may mean tools that are more integrated into the design workflow, but the concrete consequences depend on factors not yet documented, such as the openness of the announced integrations and the arrival of the low-cost variants. The full picture will require independent verification of the claimed qualities.
