Canva Craft Explained

Canva Craft is an in-house artificial intelligence (AI) image generation model created by Canva Research to generate visual assets with legible typography and structured layouts. Released in October 2026, Canva Craft is Canva's first image model trained in-house; Canva previously used a number of outside image models.

The model operates through a two-stage planning and rendering process that outputs graphics at up to 2K resolution across more than 30 aspect ratios. It is available to all Canva account holders through the Canva Dream Lab interface and to developers through the public Canva application programming interface (API).

Canva designed Canva Craft to integrate directly with Canva Magic Layers, enabling users to convert raw raster outputs into individual editable design elements. While Canva published benchmark claims from the crowdsourced DesignArena leaderboard, commercial licensing terms and per-tier consumption caps remain unpublished on canva.com.

What Canva Craft Is

Canva Craft is Canva's first in-house image generation model; it plans the layout and any text before it renders, so images come out at up to 2K with legible text and can be turned into editable Canva designs.

Canva announced the model through the Canva newsroom, describing Canva Craft as a foundational system built specifically for graphic design tasks. Canva Research, Canva's in-house AI lab, developed and trained the model from scratch specifically for design.

According to reporting from TechCrunch, Canva Craft is available immediately to all users inside the Canva Dream Lab workspace tab. Programmers can also query Canva Craft through the Canva public API to integrate generation pipelines into custom software. Canva stated that it plans to expand access to third-party conversational interfaces, such as ChatGPT and Gemini, using the Model Context Protocol (MCP) framework.

Canva says more than a quarter of a billion people use Canva, creating over 400 designs every second. TechCrunch reported that Canva acquired media generation platform Leonardo in 2024, and built Canva Craft to compete with models from Google, OpenAI, ByteDance and Meta. TechCrunch notes it likely saves API costs.

Who this is not for: Teams seeking unguided photorealistic generative art without graphic design typography should compare the Midjourney alternatives and Flux alternatives lists.

• Availability: accessible to all registered accounts in the Canva Dream Lab tab and via public API
• Third-party integrations: planned deployment to ChatGPT and Gemini through Model Context Protocol (MCP)
• Output resolution: generates images up to 2K resolution across more than 30 aspect ratios
• Corporate development: developed and trained entirely in-house by the Canva Research laboratory

Two-Stage Planning and Rendering Architecture

Canva Craft separates composition planning from visual rendering through a dual-model pipeline designed to preserve spatial coherence across graphic layouts.

Canva Craft divides generation into two distinct steps. First, a dedicated image planning model reads the prompt and drafts an explicit compositional blueprint. This structural plan defines which visual objects must appear, their spatial relationships, their relative scale, the visual hierarchy of the canvas, and the placement of any requested text strings.

Second, a specialized rendering model receives this layout plan and generates the final high-resolution raster image. Canva states that separating layout planning from surface rendering gives more legible text, precise object placement, and consistent visual relationships.

This decoupled pipeline also supports targeted adjustments to specific regions of an image. Users can modify individual objects or alter text strings without unnecessarily altering everything else. TechCrunch noted that generated outputs connect directly with Magic Layers, which turns generated images into fully editable design elements.

• Stage one planning: calculates compositional hierarchy, scale, spatial relationships, and planned typography
• Stage two rendering: executes final raster rendering from the structural blueprint at up to 2K resolution
• Local adjustments: enables isolated edits to individual objects without unnecessarily altering everything else
• Magic Layers conversion: transforms static output pixels into editable Canva elements and text blocks

Template Training Data and Designer Fine-Tuning

Canva Research trained Canva Craft on millions of human-made templates from Canva's content library.

Most image models are trained mainly on photographs. In contrast, Canva Craft was trained on millions of design templates sourced from the internal Canva library. This dataset included professionally designed posters, social posts, presentations, infographics, menus and typography-rich layouts.

Following initial pre-training, Canva Research conducted supervised fine-tuning and reinforcement learning using feedback from Canva's professional designers. Dozens of researchers and professional designers evaluated more than 25,000 image pairs during internal evaluations.

Canva reported that model distillation allowed engineers to cut inference steps to a fraction of the baseline requirement. Canva also built a custom inference engine specifically for the planning phase, making Canva Craft an order of magnitude faster than its initial internal build.

DesignArena Benchmark Rankings and Performance

Canva published performance results from DesignArena, a crowdsourced evaluation leaderboard dedicated to measuring AI graphic models.

On the global DesignArena rankings, Canva Craft debuted in fourth position overall among all evaluated image generation models. Canva says this makes Canva Research the world's second most advanced lab for image generation. The model achieved top placement in specific thematic categories while ranking lower in product imagery.

Canva reported that Canva Craft secured the number one global position in the People and Portraits category against a field of 87 competing models. The system also captured first place in the Landscapes and Nature category among 85 evaluated engines. For abstract compositions, the model ranked second, while securing fourth place on product-focused benchmarks.

Unpublished Pricing and Operating Terms

Canva has not published specific pricing schedules, usage caps, or commercial licensing terms dedicated to Canva Craft generation.

While Canva confirmed that Canva Craft is accessible to all account holders from the Dream Lab tab, Canva has not clarified whether free tier users receive the same Craft allowance as Canva Pro or Canva Teams subscribers. Rate limits and API pricing are not published. Prospective users must verify active account quotas directly on canva.com.

Canva also announced two upcoming distilled variations, named Craft-fast and Craft-ultrafast. Canva stated that these upcoming models will generate visual assets at a fraction of the cost required by leading frontier systems, but Canva has not published release dates or technical specifications for either variation. Furthermore, Canva has not stated whether Canva Craft is slated to replace older models in Dream Lab.

What would change this assessment: Canva publishing transparent monthly generation limits per subscription tier, clear commercial IP indemnity terms, and public API pricing tables would make Canva Craft easier to recommend for enterprise creative operations.

• Per-tier allowances: Canva has not published whether Free, Pro, and Teams accounts share equal access limits
• Fast model release dates: launch schedules for Craft-fast and Craft-ultrafast remain unannounced
• Commercial rights: commercial-use terms specific to Craft output are unpublished

Steps to Test Canva Craft in Dream Lab

Users can generate images with Canva Craft inside Canva.

To generate images with Canva Craft, follow these operational steps:

1. Open the Canva application and sign into an active user account.

2. Select the Dream Lab tab.

3. Enter a prompt for the graphic, including any text that must appear.

4. Use Magic Layers to turn the generated image into editable design elements.

Should Teams Switch from Existing Image Generators

Deciding whether to replace third-party generation platforms with Canva Craft depends on whether your team requires standalone artistic imagery or editable marketing production assets.

At Layer3Labs, we build and run AI systems for client workflows. Canva built Craft to compete with outside models, stating its own model provides better quality and speed control.

Teams currently operating standalone generators like Ideogram or Midjourney should review our Canva alternatives for AI image design guide. If your primary bottleneck is rendering legible text on marketing posters, Canva Craft's template-trained planning pipeline connects directly with Magic Layers for editing. For alternative typographic generators, explore our Ideogram alternatives guide.

Conversely, if your creative team depends on open-weight local execution or granular rendering controls, inspect our Flux alternatives overview. For teams evaluating presentation and slide builders, our Prezi vs Canva comparison breaks down collaborative workspace differences.

To evaluate how generative graphics impact content authentication and marketing verification workflows, review our guide to AI image detectors compared. Before migrating active production pipelines, check active account allowances and feature updates directly on canva.com.