Leonardo AI Review: 1 Billion Images In, Then Canva Bought It

By the time Canva announced it was buying Leonardo AI in July 2024, the Sydney-built app had already been used to generate more than a billion images by roughly 19 million people, according to reporting on the deal. That's not a rounding error. That's a genuinely large AI image tool getting absorbed by an even bigger design company, and two years later it still runs as its own separate product instead of getting quietly folded into Canva's toolbox.

Leonardo AI is a browser-based AI image and video generator built for people who want more control over their output than a single text box gives them. Think concept artists, game studios, product teams, and marketers who need the same character or style to show up consistently across dozens of images, not just one lucky roll of the dice. This review covers what it actually does, what it costs after a 2026 rename of its pricing tiers, how it stacks up against Midjourney and Adobe Firefly, and whether it's still worth your time now that Canva owns it.

Quick verdict

What works well and what to watch

What works well:

• A genuinely wide model lineup (Phoenix, Lucid Origin, Flux Dev/Schnell, Anime XL, Kino XL) accessible from one interface, so you pick the right engine per job instead of fighting one model's house style.

• Real creative tools beyond text-to-image: an AI-assisted Canvas for inpainting and outpainting, Motion for turning stills into short clips, a Universal Upscaler that goes up to 8K, and 3D texture generation for game assets.

• Paid plans grant full commercial ownership of your outputs, not just a usage licence.

• A clean, guided web interface. No Discord server required to get started, unlike Midjourney.

What to watch:

• The token system takes real effort to understand, and "unlimited" generations only apply to Leonardo's own first-party models. Premium third-party models like Veo, Kling, and Sora still burn tokens at full rate even on the top plan.

• Free-tier images are public by default and Leonardo keeps the IP rights to them, so casual users don't actually own what they make until they pay.

• Its public affiliate program is closed to new applicants as of April 2026 (more on that in the FAQ below), so if you were hoping to promote it for commission, that door is not open right now.

What is Leonardo AI, really?

Leonardo AI launched publicly in December 2022, founded in Sydney by JJ Fiasson and a small team of co-founders, several with backgrounds in gaming. It grew fast: about 7 million users generating 4.5 million images a day within its first year, on the way to the billion-image mark noted above. Investors including Blackbird and Samsung Next backed it with roughly AU$47 million before Canva acquired the company outright in July 2024.

Since the acquisition, Leonardo has kept operating as its own product, with its own pricing, its own app, and its own release schedule, while its underlying Phoenix model also powers Canva's Dream Lab and Magic Media generators inside Canva itself. So you are not just testing a standalone image generator here, you are testing the same engine that now sits behind one of the biggest design platforms on the internet. Canva has said that Magic Studio, the suite Leonardo's tech now feeds, handles billions of AI-assisted actions across Canva's own user base of roughly 190 million people, which gives you a sense of scale: a niche indie image generator two years ago is now running underneath one of the most-used design apps on the planet.

Key features

• Phoenix. Leonardo's own foundational model, built for prompt adherence and legible in-image text, with fast, quality, and ultra generation modes so you can trade speed for fidelity depending on the job.

• Lucid Origin. A newer, more general-purpose model aimed at a wide stylistic range without losing prompt accuracy, useful when Phoenix feels too locked into one photographic look.

• Realtime Canvas. A single workspace for inpainting, outpainting, and compositing, closer to a lightweight Photoshop than a one-shot prompt box.

• Motion. Turns a still image or prompt into a short animated clip, with basic camera controls like pan, zoom, and tilt, useful for product shots and social content.

• Universal Upscaler. Pushes finished images up toward 8K resolution using generative refinement rather than simple pixel stretching.

• 3D texture generation. Upload a 3D mesh and Leonardo paints albedo, normal, and roughness texture maps onto it directly, ready to drop into Unity or Unreal. Arguably its most unusual and least copied feature.

• Custom model training. Train the platform on your own reference images to lock in a specific character, product, or art style across every generation, the feature that separates a one-nice-image tool from an actual production pipeline.

• API access. Pay-as-you-go, starting from a small starter credit balance, for teams that want to build Leonardo generation into their own app or workflow.

How it performs

We did not run a paid Leonardo AI subscription for this review, so treat the following as grounded in Leonardo's own documentation and published specs rather than a first-hand production test. Verify anything mission-critical against your own trial before committing budget.

What is clear from the documentation alone is that Leonardo's real edge is not any single image, it is the workflow built around the image. A Midjourney user gets one strong aesthetic and iterates on a prompt. A Leonardo user picks a model for the job, generates in Canvas, upscales the keeper, trains a custom model on it if they need the same character again next week, and exports straight into a game engine if it is a texture. That is a materially different, more production-oriented use case than making a pretty picture, and it is why Leonardo shows up so often in game-dev and concept-art circles specifically, rather than general social content.

Leonardo's Iterate feature is a small detail worth calling out on its own: it lets you nudge a finished generation with short, plain-English phrases, such as "make the lighting warmer" or "move the character left," instead of rewriting your whole prompt from scratch and hoping the model remembers everything else you liked. It sounds minor until you have used a tool without it, at which point it becomes the difference between thirty seconds of editing and fifteen minutes of prompt archaeology.

The free plan is genuinely usable for testing this out. At 150 tokens a day, with no rollover, that is enough for roughly 75 to 150 generations depending on the model and settings you pick, more breathing room than most rivals give away for nothing.

Pricing

Leonardo renamed its individual plans in 2026 (Apprentice, Artisan, and Maestro became Essential, Premium, and Ultimate), but the prices and token allowances stayed the same. Always check leonardo.ai/pricing directly before buying, since AI tool pricing shifts often.

All annual plans save roughly 20 percent over paying monthly, which is standard for the category. Every paid tier, Essential and up, also swaps the free plan's daily reset for a monthly balance that rolls unused tokens forward, which matters if your workload is bursty rather than steady, a product launch week versus a quiet month, for example.

Is it worth the price? For Essential and up, yes, if you will actually use the private generations and model training. The Free plan is fine for kicking the tyres, but the moment you want to own an image or keep it off the public feed, you are paying. That is true of almost every serious AI image tool at this point, not just Leonardo.

Leonardo AI vs the competition

None of the tools below are direct clones of each other, so the honest comparison is less about which one is best and more about which one matches how you actually work.

If style and mood matter most to you, our Midjourney review is worth reading side by side with this one. If you are already inside Adobe's ecosystem, our Adobe Firefly review and, since Firefly lives inside it, our Adobe Photoshop review cover that path. And if all you actually want is a finished poster or social post without touching a model picker, our Canva review covers the tool that now literally runs Leonardo's own tech under the hood.

Who should use Leonardo AI

Not every AI image tool needs to do everything well. Here is where Leonardo actually earns its subscription, and where it does not.

Best for:

• Concept artists and game studios who need production-ready, UV-mapped textures and consistent characters.

• Marketers and small teams who want to train a model on their own brand or product assets.

• Anyone who wants more knobs to turn than a single text box, without learning Discord commands to get there.

Skip it if:

• You just want one great-looking image occasionally and do not want to think about tokens. Canva's own free AI tools will get you there faster.

• You need airtight commercial indemnification for enterprise client work. Adobe Firefly's terms are built specifically for that.

• You were hoping to earn commission promoting it. The affiliate program is not accepting new applicants right now.

Final verdict

Leonardo AI earned its spot as one of the few AI image tools that survived an acquisition by a much bigger company and kept its own identity. Canva could have quietly retired the brand and just kept the Phoenix model for itself. Instead, two years on, Leonardo still ships its own features, at its own pace, at leonardo.ai.

If you need control, consistency, and production-ready assets rather than one lucky prompt, it is genuinely worth the $12-a-month entry price. If you just want a nice graphic for a slide deck, you may not need to leave Canva at all, which, funnily enough, now runs on the same engine underneath.

One honest note before you click anywhere: this recommendation is not riding on a commission. Leonardo AI's public affiliate program stopped accepting new applicants on April 7, 2026, so there is no AIwithKay partner link doing any of the talking here, just a straightforward look at whether the tool is worth your money on its own terms.