AI Image Generator Test: Canva Craft vs. ChatGPT, Gemini, Meta Muse

Canva hopes to win a contest for artificial intelligence image generation.

design platform On Thursday it launched Canva Craft, its first image generation model, similar to those from Google, Meta and OpenAI.

Users can access Craft through Canva’s Dream Lab, including with a free account, but it is subject to AI usage restrictions. You need to log in to your Canva account to get started.

Canva said in the launch announcement that the tool plans the layout, proportions and text of a design before generating the image, an approach designed to help it produce crisp fonts and accurate detail.

The company also said the model was trained on human-designed Canva templates, including posters, menus and presentations, rather than photos.

To test Canva’s capabilities, Business Insider created three image generation prompts with increasing levels of difficulty—from designing a simple greeting card to creating realistic images from scratch.

Here’s how Canva’s new image generation tool compares to three of the most powerful models.

Canva did not respond to a request for comment.

Start with illustrations

We started simple, prompting for a birthday card.

Create a cartoon “Happy Birthday” card for a friend named Jessica. Jessica has pink hair, brown eyes, likes paisley patterns, and will be 28 years old in November this year.

Surprisingly, design platform Canva falls short when it comes to birthday card designs, from clunky fonts to oversized cartoon character heads. It didn’t even bother to include any paisley patterns.

In contrast, ChatGPT and Muse understand this task.

ChatGPT guessed that the recipient’s favorite color might be pink, and Muse thoughtfully incorporated cozy fall elements for the girl with a November birthday.

Meanwhile, Gemini took the request literally and produced an image of the actual card. Bonus points!

Winner: Muse

Edit image

For the second task, I tested the AI ​​model’s ability to identify elements and place them in a novel environment.

Edit this picture of me holding the necklace pendant. Take the red garnet pendant off my hand and make it look like it’s on display in a glass case in a jewelry showroom.

Canva does a great job of identifying pendants and creating atmosphere, but like Gemini, it depicts the jewelry in a way that defies physics and common sense. It is impractical to have a pendant floating in the air or supported on a metal pole.

ChatGPT and Muse again outperformed other models in understanding how things typically look in reality. However, Muse made some changes to the pendant that I did not ask for.

Winner: ChatGPT

Create photorealistic images from scratch

With Halloween quickly approaching, I asked for realistic depictions of my favorite animals.

Produces a realistic image of a long-haired orange tabby cat with bat wings wearing a Halloween costume, perched on the edge of a gothic medieval castle under a full moon and starless night sky.

While the cat’s fur looks a little too fluffy and high-contrast in Canva’s creation, no components are missing or misshapen, and the image composition matches the quality of Muse and Gemini quite well.

However, for me, the clear winner of this round will be ChatGPT.

Not only is the cat’s pose majestic, but the soft glow on the left is extremely realistic and the white balance is well done. There are also clouds gently holding the moon, which may be a response to my request for a starless sky.

Winner: ChatGPT

Overall, it’s clear that Canva still has a lot of room to grow compared to other models, especially if its focus is on graphic design. For now, I’ll stick to moving the components manually.