Canva Craft review: 2K resolution limits, API pricing and GPT Image alternatives

Canva Craft launched on 22 October 2026 as the company’s first in-house AI image model. It produces images up to 2K resolution and performs best on design layouts rather than pure photography. The model integrates directly with Canva’s editor and the Connect API.

Users access Craft through the Dream Lab tab. It consumes existing monthly AI allowances instead of separate per-image fees. Free accounts receive roughly 20 generations per month, Pro plans allow up to 200 Premium uses, and Business plans reach 400. API calls also debit the same allowance pool.

Canva Craft review: technical limits and API access

The model uses a two-stage process that first plans composition then renders the image. It supports more than 30 aspect ratios plus custom sizes and generates legible text in 55 languages. Training data came mainly from Canva templates, which explains its strength in posters, infographics and typography-heavy designs.

Via the public API, developers call the endpoint with the craft_v1_2k model. Rate limits sit at 20 requests per minute per user token. Each job takes 50–60 seconds and returns a download URL valid for 24 hours. The feature remains in preview and requires the asset:write scope.

DesignArena rankings placed Craft at #4 overall on launch day with an ELO score between 1,329 and 1,370. It ranked first in People & Portraits and Landscapes & Nature categories but fourth in Products. The model currently sits behind three OpenAI GPT-Image-2.5 variants on the same board.

Canva Craft vs GPT Image and Google alternatives

Compared with GPT Image 2.5, Craft shows clearer text placement and better adherence to design hierarchy, yet it lags in photorealism and maximum resolution. Google’s Nano Banana 2.1 offers standalone API pricing from $0.03 to $0.11 per image depending on resolution, while Craft has no separate charge beyond the user’s Canva allowance.

Microsoft and Meta models rank below Craft on early DesignArena votes. For workflows that stay inside Canva, the direct export to editable Magic Layers provides a practical advantage over external generators that require file import and manual layer recreation.

Upcoming faster variants and MCP server support are scheduled but not yet live. Users who need 4K output or pure photographic quality still turn to OpenAI or Google services for those specific tasks.