Leonardo AI Review: A Versatile Image Generator for AI Creative Enthusiasts – CNET [CNET]

View Article on CNET

Leonardo AI

Pros

  • Clear and detailed images
  • Helpful prompt engineering tools

Cons

  • No post-generation editing tools on the free plan
  • Lackluster privacy policy

Not da Vinci or DiCaprio, but still within the realm of artistic Leos, Leonardo AI is an image generator. It was originally meant to help folks create gaming assets, but it’s now a full-scale AI content creation service, with AI video creation and editing services in addition to image tools.

ai-atlas-tag.png

Overall, Leonardo is a solid option compared with its many AI-powered competitors. It’s on par with Adobe Firefly and miles ahead of Google’s ImageFX and Canva. Leonardo adheres to prompts better than Midjourney but lacks extensive editing tools, making it a toss-up between the two. OpenAI’s Dall-E 3 continues to be CNET’s highest-ranked pick, but you’ll need to cough up $20 for ChatGPT Plus, whereas Leonardo has a comprehensive free plan.

An AI rendering of Da Vinci's Last Supper

I had to try to recreate The Last Supper. Fortunately for da Vinci, his mastery remains unbeatable, not in the least because Leonardo AI only included eight disciples.

Created by Katelyn Chedraoui using Leonardo AI

I generated over 90 images with Leonardo, from stock images to sci-fi and fantasy renderings. Here’s how it went.

How CNET tests AI image generators

CNET takes a practical approach to reviewing AI image generators. Our goal is to determine how good it is relative to the competition and which purposes it serves best. To do that, we give the AI prompts based on real-world use cases, such as rendering in a particular style, combining elements into a single image and handling lengthier descriptions. We score the image generators on a 10-point scale that considers factors such as how well images match prompts, creativity of results and response speed. See how we test AI for more.

Leonardo can use your content to “improve its services” with machine learning, which essentially means it can train on your images, according to its terms of service. If you don’t want your images to be public, you’ll need to invest in a paid plan. Leonardo automatically sets you up on its free plan when you sign up. With the free plan, you get 150 tokens every eight hours, which roughly translates to 13 to 15 batches of images. Additional editing tools are available with the paid premium plans and may cost you more tokens to use.

How good are the images, and how well do they match prompts?

Da Vinci probably would’ve scoffed at the idea of AI-generated art, but I was pretty happy with the images Leonardo produced. The clarity of Leonardo’s images stood out to me compared to those from other generators. Its details were finely rendered, and, judging by one of my personal metrics for cleanliness, most hands had a normal number of fingers. Occasionally the size of certain elements would feel disproportionate, but it would just give the image more of a whimsical vibe rather than be totally unusable.

Leonardo is more likely to match an overall vibe for a prompt rather than adhere to a list of individual elements. Meaning, if you want a cinematic scene of a dragon flying over a castle, you’ll be pleased with the results. But when I specifically asked for a big black dragon and a smaller golden one above a castle with four turrets and a red flag waving, I didn’t get exactly what I had envisioned.

a big black dragon standing in front of a castle with waving red flags

I listed out specific characteristics for the dragons and the castle, but I only got a few.

Created by Katelyn Chedraoui using Leonardo AI

How engaging are the images?

Leonardo’s images were very engaging. Its level of detail and crisp colors made images fun to create and inspect. 

Because Leonardo’s images are so engaging, you might be tempted to try out some of Leonardo’s other AI creator tools like Canvas Editor and Realtime Generation. But I recommend skipping them. Those programs are much less user-friendly and the content created was much worse, with blurry, off-center or weird quirks. There is much better image editing software available, and Meta AI’s “Imagine” function is a more accurate real-time image generation tool.

Can you fine-tune results?

Leonardo does not offer post-generation editing tools for individual images unless you have a premium plan. If you’re on the free plan, you’ll need to submit a new, edited prompt to make adjustments. The best way to avoid that is to take advantage of all of Leonardo’s customization options and build out the best possible prompt.

Every prompt should have a preset style, contrast level and dimension selected from the left-hand menu. For more advanced prompts, you can use Image Guidance and Elements. Using the star icon in the prompt box, Image Guidance lets you add one of your previous images to the new prompt as a style or color reference. Elements work similarly, letting you choose specific predetermined looks like glassy and soft pastel anime. Tacking those extra elements onto your prompts can help steer the AI in the right direction.

Leonardo’s paywalled Alchemy Refiner promises to “refine and upscale” images, particularly for faces and hands, which are notoriously tricky for AI image generations. Because I was on the free plan, I couldn’t test this myself, but I was notably impressed by the clarity and accuracy of people’s hands and teeth compared to those from other AI generators.

two cowboys locked in a fight embrace, another looks onward in an old Western town

I have very low expectations of AI-generated hands and teeth, but Leonardo’s are pretty good.

Created by Katelyn Chedraoui using Leonardo AI

Leonardo offers two great features to help you with your prompt engineering: a random prompt creator and a prompt improvement tool. I found both features helpful for sparking inspiration and guiding Leonardo toward my vision. Using the Improve Prompt tool, I entered a simple prompt, and Leonardo built out that idea into a longer paragraph that addressed multiple objects and the overall vibe. If you get stuck, the random prompt generator is good for brainstorming. 

How fast do images arrive?

Images arrive within 10 to 20 seconds, which makes Leonardo one of the faster AI image generators. Image generation time can vary depending on the model you use; the new Phoenix model takes a bit longer, for example. But even with Phoenix I didn’t feel the need to start scrolling on my phone or check my email while waiting for the images to load, like I did with other generators.

Leonardo is nice, but I’m not wowed

Leonardo checks a lot of important boxes for AI creators. It’s fast, has a free plan and creates pretty normal-looking images. But there are a couple of things stopping me from emphatically recommending it to everyone. Paywalling post-generation editing tools is annoying, and it means you’ll run through your tokens faster trying to get what you want. The important parts of the privacy policy are hidden in the terms of service, leaving much to be desired there. And less quantifiably, I just wasn’t wowed by the results. It felt very middle of the road. Of course, there’s nothing wrong with that — it just won’t dethrone our current top pick. If you’re a non-professional creator or AI creative enthusiast, Leonardo will be a great fit for quickly and easily generating usable, albeit not perfect, AI images.

CNET is using an AI engine to help create a handful of stories. Reviews of AI products like this, just like CNET’s other hands-on reviews, are written by our human team of in-house experts. For more, see CNET’s AI policy and how we test AI.