Nipun Goyal

Engineering
May 1, 2025

Nipun Goyal

May 1, 2025
Engineering

A Deep Dive Into Canva’s New Canva Code

We’re entering a new era in which building software isn’t just the job of engineers, it’s becoming accessible to creators, designers, and marketers.

AI is redrawing the boundaries of who gets to build. Replit is turning code suggestions into full-stack scaffolds. Figma’s AI lets you create prototypes just by typing what you want. And now, Canva - long known as the “design tool for everyone”- is moving into code.

This month, Canva launched Canva Code - a new feature that doesn’t just generate HTML and CSS from your designs but also lets you create interactive widgets and small apps using plain language.

But what does that mean for the dev world? Let’s unpack.

Canva’s Most Ambitious Move Yet

We’re entering a new era in which building software isn’t just the job of engineers, it’s becoming accessible to creators, designers, and marketers.

AI is redrawing the boundaries of who gets to build. Replit is turning code suggestions into full-stack scaffolds. Figma’s AI lets you create prototypes just by typing what you want. And now, Canva - long known as the “design tool for everyone”- is moving into code.

This month, Canva launched Canva Code - a new feature that doesn’t just generate HTML and CSS from your designs but also lets you create interactive widgets and small apps using plain language.

But what does that mean for the dev world? Let’s unpack..

Canva’s Most Ambitious Move Yet

With over 170 million users globally, Canva has spent the last decade simplifying design. But now, it’s making an even bolder move.

Canva isn’t new to web building. They’ve long allowed users to create simple websites. But Canva Code is a leap forward. It takes the drag-and-drop philosophy and combines it with prompt-driven coding to output actual logic-driven components.

Example:

Type “Create a flashcard app for biology terms,” and Canva will generate code for HTML, CSS, and JS with a preview.

It’s not just visual prototyping anymore. It’s functional UI generation with the option to:

  • Edit the UI visually
  • View and tweak the code directly
  • Embed the result anywhere in your Canva ecosystem

This opens the door for non-developers to build interactive elements that would've otherwise required engineering bandwidth.

What’s Really Powering It?

Canva Code is more than a simple prompt-to-HTML tool. It’s a sophisticated AI-powered feature designed to turn natural language requests into interactive, embeddable widgets and mini-apps directly within Canva’s design environment.

Here’s what’s powering it:

  • LLM-Driven Logic Generation:

    It leverages advanced LLMs, notably Anthropic’s Claude, to interpret user prompts and generate JavaScript logic blocks for interactive features like calculators, quizzes, and timers. Users simply describe what they want, and the AI translates that into functional code, eliminating the need for manual coding.
  • Visual-to-Code Transformers:

    The system uses proprietary AI models trained on extensive design-to-markup datasets. These models analyze layout structures, spatial relationships, paddings, alignments, and styles. Then generate clean, semantic HTML and CSS.

    The resulting code closely resembles what experienced front-end developers would write, ensuring the output is both visually accurate and technically sound.
  • Componentized Architecture:

    Internally, Canva Code breaks down designs into modular, reusable components. Each component is styled and behaves independently, making use of encapsulation techniques (such as shadow DOM) to prevent style conflicts and ensure portability.

    This modularity allows widgets and mini-apps to be easily integrated into any Canva project, from presentations to websites.
  • Real-Time Code Syncing:

    As users make changes to their designs, Canva Code updates the underlying code in real time. This is achieved through a custom bidirectional sync engine that keeps the visual representation and the codebase in perfect alignment.
  • Output Format Support:
    • HTML5 (semantic, production-ready)
    • CSS3 (responsive and modularized)
    • JavaScript (vanilla, scoped to component logic only)
    • Embed script (Canva-hosted runtime with sandboxing)

This setup turns Canva into one of the very few tools where non-technical users can design, generate, and iterate on fully functional frontend widgets.

Where Canva Code Falls Short?

As advanced as Canva Code is, it’s still fundamentally a front-end code generator. It doesn’t (yet) step into the territory of product engineering or backend orchestration.

Here’s what you won’t get:

  • Backend logic or APIs: No Node.js, no data binding, no serverless functions. There’s no provision to wire up REST or GraphQL endpoints.
  • Database or state management: You can’t persist data, manage local/global state, or integrate with real-time systems like Firebase or Supabase.
  • User management: No sign-up/login flows, auth tokens, or role-based access.
  • Infra support: No hosting logic, domain linking, CI/CD, versioning, or observability baked in.
  • SDK/plugin ecosystem: There’s no way to plug in third-party services like Stripe, Segment, Chat SDKs, or analytics tools, at least not natively.

In other words, Canva Code gets you a good-looking, semi-functional UI shell, but the moment you need logic, data, or integration, you’re reaching for your IDE.

Ideal Use Cases:

  • Marketing widgets (embedded calculators, quizzes)
  • Interactive prototypes for client presentations
  • Landing pages with basic forms/animations
  • Social media embeds (polls, mini-games)

And these are the gaps that the next wave of dev tools needs to fill: AI-native builders that go far beyond UI. Good news - we at LikeMinds are building it as we speak.

We are creating an AI agent that integrates, customizes, and maintains SDKs/libraries via simple prompts and live previews - unlocking a faster way to build non-core features like chat, auth, payments, and search. And we would love to see what your views on it.

Supercharge your retention with in-app social features

Deploy customised features on top of chat and feed in 15 minutes using LikeMinds SDK.

Schedule a demo!
Image showing Resource Library, which is a inApp feed usecase.

Share now

NEWSLETTER

Get a front row seat to everything happening at LikeMinds including some curated expert insights each week, delivered straight to your inbox.
We promise to not spam. 🤝🏻

Sign Up Now!

Follow us

A Deep Dive Into Canva’s New Canva Code

Nipun Goyal
/
May 1, 2025
/

We’re entering a new era in which building software isn’t just the job of engineers, it’s becoming accessible to creators, designers, and marketers.

AI is redrawing the boundaries of who gets to build. Replit is turning code suggestions into full-stack scaffolds. Figma’s AI lets you create prototypes just by typing what you want. And now, Canva - long known as the “design tool for everyone”- is moving into code.

This month, Canva launched Canva Code - a new feature that doesn’t just generate HTML and CSS from your designs but also lets you create interactive widgets and small apps using plain language.

But what does that mean for the dev world? Let’s unpack.

Canva’s Most Ambitious Move Yet

We’re entering a new era in which building software isn’t just the job of engineers, it’s becoming accessible to creators, designers, and marketers.

AI is redrawing the boundaries of who gets to build. Replit is turning code suggestions into full-stack scaffolds. Figma’s AI lets you create prototypes just by typing what you want. And now, Canva - long known as the “design tool for everyone”- is moving into code.

This month, Canva launched Canva Code - a new feature that doesn’t just generate HTML and CSS from your designs but also lets you create interactive widgets and small apps using plain language.

But what does that mean for the dev world? Let’s unpack..

Canva’s Most Ambitious Move Yet

With over 170 million users globally, Canva has spent the last decade simplifying design. But now, it’s making an even bolder move.

Canva isn’t new to web building. They’ve long allowed users to create simple websites. But Canva Code is a leap forward. It takes the drag-and-drop philosophy and combines it with prompt-driven coding to output actual logic-driven components.

Example:

Type “Create a flashcard app for biology terms,” and Canva will generate code for HTML, CSS, and JS with a preview.

It’s not just visual prototyping anymore. It’s functional UI generation with the option to:

  • Edit the UI visually
  • View and tweak the code directly
  • Embed the result anywhere in your Canva ecosystem

This opens the door for non-developers to build interactive elements that would've otherwise required engineering bandwidth.

What’s Really Powering It?

Canva Code is more than a simple prompt-to-HTML tool. It’s a sophisticated AI-powered feature designed to turn natural language requests into interactive, embeddable widgets and mini-apps directly within Canva’s design environment.

Here’s what’s powering it:

  • LLM-Driven Logic Generation:

    It leverages advanced LLMs, notably Anthropic’s Claude, to interpret user prompts and generate JavaScript logic blocks for interactive features like calculators, quizzes, and timers. Users simply describe what they want, and the AI translates that into functional code, eliminating the need for manual coding.
  • Visual-to-Code Transformers:

    The system uses proprietary AI models trained on extensive design-to-markup datasets. These models analyze layout structures, spatial relationships, paddings, alignments, and styles. Then generate clean, semantic HTML and CSS.

    The resulting code closely resembles what experienced front-end developers would write, ensuring the output is both visually accurate and technically sound.
  • Componentized Architecture:

    Internally, Canva Code breaks down designs into modular, reusable components. Each component is styled and behaves independently, making use of encapsulation techniques (such as shadow DOM) to prevent style conflicts and ensure portability.

    This modularity allows widgets and mini-apps to be easily integrated into any Canva project, from presentations to websites.
  • Real-Time Code Syncing:

    As users make changes to their designs, Canva Code updates the underlying code in real time. This is achieved through a custom bidirectional sync engine that keeps the visual representation and the codebase in perfect alignment.
  • Output Format Support:
    • HTML5 (semantic, production-ready)
    • CSS3 (responsive and modularized)
    • JavaScript (vanilla, scoped to component logic only)
    • Embed script (Canva-hosted runtime with sandboxing)

This setup turns Canva into one of the very few tools where non-technical users can design, generate, and iterate on fully functional frontend widgets.

Where Canva Code Falls Short?

As advanced as Canva Code is, it’s still fundamentally a front-end code generator. It doesn’t (yet) step into the territory of product engineering or backend orchestration.

Here’s what you won’t get:

  • Backend logic or APIs: No Node.js, no data binding, no serverless functions. There’s no provision to wire up REST or GraphQL endpoints.
  • Database or state management: You can’t persist data, manage local/global state, or integrate with real-time systems like Firebase or Supabase.
  • User management: No sign-up/login flows, auth tokens, or role-based access.
  • Infra support: No hosting logic, domain linking, CI/CD, versioning, or observability baked in.
  • SDK/plugin ecosystem: There’s no way to plug in third-party services like Stripe, Segment, Chat SDKs, or analytics tools, at least not natively.

In other words, Canva Code gets you a good-looking, semi-functional UI shell, but the moment you need logic, data, or integration, you’re reaching for your IDE.

Ideal Use Cases:

  • Marketing widgets (embedded calculators, quizzes)
  • Interactive prototypes for client presentations
  • Landing pages with basic forms/animations
  • Social media embeds (polls, mini-games)

And these are the gaps that the next wave of dev tools needs to fill: AI-native builders that go far beyond UI. Good news - we at LikeMinds are building it as we speak.

We are creating an AI agent that integrates, customizes, and maintains SDKs/libraries via simple prompts and live previews - unlocking a faster way to build non-core features like chat, auth, payments, and search. And we would love to see what your views on it.

Supercharge your retention with in-app social features

Deploy customised features on top of chat and feed in 15 minutes using LikeMinds SDK.

Let's start!