Leonardo.ai

What it actually does

Leonardo.ai generates images from text prompts, with a library of models fine-tuned for particular looks (anime, photoreal, fantasy, 3D render). It also does image-to-image guidance, an AI canvas for editing, 3D texture generation, and training on your own dataset. The Real-Time Canvas renders as you draw, which is the feature people actually come back for.

Ownership matters here. Canva announced its acquisition of Leonardo on 29 July 2024, and all of roughly 120 Leonardo staff joined Canva. Startup Daily reported the deal at at least around $320 million. The standalone product still runs, but its long-term future is now a Canva roadmap decision, and that worry comes up repeatedly in Leonardo's own user community.

What it costs

The free plan gives 150 images a day and 30 refined images a day, which is unusually generous for this category. Apprentice is $12 a month, Artisan is $30, Maestro is $60.

The thing buyers miss is the credit economy underneath those tiers. Heavy users burn through tokens faster than they expect and end up paying more than the headline price implies, which is the same complaint the free plan generates when credits run dry mid-session. The other cost surprise is the renewal. Reviewers report mistaken or unwanted charges going through on a mis-click, and auto-renewals catching people who thought they had stopped paying. Refund handling is the most-discussed friction point in public reviews of this product.

Where it falls down

Billing confusion is the recurring complaint, not image quality. Users describe charges they did not intend to make, then chasing a refund afterward. If you subscribe, note the renewal date somewhere you will see it.

Content and safety filters are the second theme. Users report benign prompts getting blocked with no clear reason, which is irritating when you are iterating on a paid plan and each attempt costs credits.

Support has a split history. There is a long-standing complaint that paying customers had effectively no support channel, though recent Trustpilot reviews are noticeably more positive and repeatedly single out agents as fast and effective at resolving charges. We would call that improved rather than fixed.

On output, artistic quality sits below Midjourney for fine art work, and the interface is dense enough that new users find it overwhelming. Neither is disqualifying, but if you are chasing the best single image rather than volume and control, you will notice.

The public scores are better than most tools in this category: 4.6 on Trustpilot across 1,465 reviews, and 4.5 on G2 across only 32, where the seller is now listed as Canva. That G2 sample is too small to lean on. Our 3.5 is the Trustpilot picture discounted for the billing friction and for the ownership question hanging over the standalone product.

Who should buy it

Game developers and content creators who need a lot of images with real control over style should try it, and the free plan is large enough to make that trial meaningful before any money moves. If you are producing game assets or iterating on a look across dozens of variations, the fine-tuned model library and the guidance controls are worth the $12 tier.

Do not buy it if you want the highest artistic ceiling per image. Midjourney is still better for that, and you should pay for it instead. Do not buy it if you need a tool with a guaranteed independent future, because Canva owns the roadmap now. And do not buy it if you dislike credit meters, because the token economy is where the advertised price and the real price separate. Anyone who wants unlimited local generation with no per-image accounting should look at running open weights on their own hardware instead.