Leonardo AI Review: Powerful Image Generator, Painful Billing

Overview

Leonardo.ai is what happens when an image generator is built by people who needed the images for something. Founded in Australia in 2022 by Jachin (JJ) Fisher and team — originally to solve their own problem of producing game assets at indie-studio speed — Leonardo grew into one of the most-used AI image platforms in the world, attracted tens of millions of registered users, and in 2024 was acquired by Canva, which now runs it as its dedicated pro-grade generative arm while folding its technology into Canva's own products. That origin story explains the product's entire personality: where Midjourney optimized for the single stunning image, Leonardo optimized for production workflows — consistency, control, volume, and the unglamorous plumbing (transparent backgrounds, tiling textures, reference-guided poses, trainable custom styles) that separates "pretty picture" from "usable asset."

The platform today is a full creative workbench in the browser, no Discord required. At its center sit Leonardo's own model families — the current Lucid generation and the Phoenix line before it — trained in-house and tuned for prompt adherence, coherent text rendering, and graphic-design-friendly output, alongside licensed heavyweights like the FLUX models and a deep catalog of community fine-tuned models specialized for everything from anime to isometric game tiles. Around the generators wraps the toolset that earns the "workbench" label: Elements for mixing style ingredients like a palette; a Consistent Character engine for keeping the same face and figure across scenes; custom model and LoRA training, where you teach the platform your own style or product from a handful of images; ControlNet-style guidance for dictating pose, depth, and composition from reference images; a Canvas editor for inpainting, outpainting, and assembling; Realtime Canvas, which redraws as you sketch; upscalers, background removal, and transparent-PNG generation for assets that drop straight into engines and design tools; and Motion for animating stills into short clips. An API and team plans extend the same machinery to studios and products.

The economics run on tokens, and the system rewards attention. Every generation costs tokens scaled by model, resolution, and features — a standard image runs a handful, premium models and enhancement passes multiply the burn — and the tiers are really token allowances with feature gates. Two wrinkles matter more than the numbers. First, Leonardo has moved toward "unlimited" generation on its own in-house models at paid tiers while metering premium and partner models by token — generous in spirit, and confusing in practice, since the models you most want for a given job may be the metered ones. Second, and underappreciated: the free tier's ownership terms. Free generations are public, and the fine print grants Leonardo broad rights over them — you keep a commercial license to use what you make, but you don't hold the IP the way paid subscribers do, whose private generations remain fully theirs. For hobbyists this is trivia; for anyone producing anything they care about owning, it's the single best reason the $12 tier exists.

One positioning note professionals should have on file: Leonardo occupies the pragmatic middle of the category's legal spectrum. It doesn't offer the licensed-training-data guarantees and enterprise indemnification that make Adobe Firefly the legal department's darling, nor does it carry the open-web-scraping baggage in its rawest form — its terms grant paid users clean ownership of outputs, its moderation has tightened considerably from the platform's looser early days, and its model catalog mixes in-house training with licensed partners. For the overwhelming majority of commercial use — marketing assets, game art, content imagery — that's ample. For the narrow slice of buyers whose counsel demands provenance paperwork and indemnities, Firefly remains the category's only full answer, and Leonardo doesn't pretend otherwise. The platform's scale, meanwhile, has crossed the thresholds that matter for longevity: tens of millions of users, billions of cumulative generations, polished mobile apps alongside the web workbench, and — post-acquisition — the balance sheet of one of software's most successful private companies behind it.

Where does Leonardo sit in the 2026 image wars? Honestly positioned: its native models are very good and occasionally excellent, but the pure-aesthetics crown still tends toward Midjourney, and the best partner models inside Leonardo's own catalog often outdraw its house engines — the same awkward dynamic Adobe faces with Firefly. Leonardo's answer is the same too, and it's a good one: compete on the workflow, not the single image. For game developers, marketing teams, print-on-demand sellers, concept artists, and anyone whose need is "many controlled, consistent images that fit a pipeline" rather than "one masterpiece," the toolset advantage is real and the pricing undercuts most professional alternatives. Meanwhile the Canva acquisition has cut both ways: resources, stability, and enterprise legitimacy flowed in, while some early power users grumble that the platform's indie edge — the rapid experimental model drops, the hobbyist-first energy — has mellowed into something more corporate. Both observations are accurate; which one matters depends on whether you're here to tinker or to ship.

Pricing & Plans

Leonardo's tiers are token allowances with feature gates, discounted roughly 17–20% on annual billing — and the most consequential line isn't a price but a term: free-tier images are public with limited ownership, while paid tiers grant private generation and full IP rights.

Two buying notes: the free tier's daily reset makes it a genuinely sustainable sandbox — unusual in this category — and the Apprentice upgrade is less about tokens than about the three unlocks (privacy, IP, character consistency) that convert the platform from toy to tool.

Key Features & Capabilities

• In-house model families plus a licensed catalog — Leonardo's own Lucid/Phoenix models deliver strong prompt adherence and clean graphic output, flanked by licensed frontier models and thousands of community fine-tunes specialized by style and subject. The practical win is choice-per-task inside one interface — with the honest caveat that the house models aren't always the best models on their own shelf.
• Custom model and LoRA training — Upload a modest set of images and train a model on your art style, product line, or character — then generate endlessly within it. This is Leonardo's deepest professional feature: brands lock visual identity, game studios lock asset styles, and artists scale their own hand. Included from the first paid tier, where competitors gate it far higher.
• Consistent Character engine — Keep the same face, figure, and outfit across poses, scenes, and expressions — the feature that unlocks storyboards, comics, brand mascots, and children's-book pipelines that single-shot generators can't sustain. Not flawless, but among the best implementations at consumer prices.
• Control tools (Elements, image guidance, ControlNet) — Mix style ingredients with sliders, steer composition from reference images, and dictate pose, depth, or line art precisely. This control stack is what "production workflow" means in practice: the image you specified, not the image the dice rolled.
• Canvas and Realtime Canvas — In/outpainting, compositing, and touch-up in a proper editor, plus a real-time mode that renders as you sketch — genuinely useful for concepting and one of the more fun onboarding experiences in AI art.
• Asset-pipeline utilities — Transparent PNG generation, background removal, high-quality upscaling, and texture-friendly output that drop straight into game engines, design tools, and print workflows. Unsexy, decisive: this is the trench where Leonardo beats prettier rivals for working users.
• Motion (image-to-video) — Animate generated or uploaded stills into short clips. Serviceable for social-format life and previz, and honestly a tier below the dedicated video platforms — think finishing touch, not video strategy.
• Community and templates — A vast public feed of images with visible prompts and settings, remixable in one click — simultaneously the best prompt-education resource on the platform and the reason free users should remember their generations are part of it.

Verdict

Our verdict sorts by what you're making, because Leonardo is a specialist wearing a generalist's interface.

The strongest yes goes to production creators: game developers first among them — the platform's original DNA still shows in texture, sprite, and concept workflows — followed by marketing and e-commerce teams generating controlled product and campaign imagery, print-on-demand and content sellers who need volume with consistency, and illustrators building repeatable characters and styles. For all of these, the combination of custom training, character consistency, control tools, and pipeline utilities delivers something Midjourney structurally doesn't: images as components of a system rather than outputs of a slot machine. At $12–30 a month with full IP rights from the first paid tier, the value against professional alternatives is excellent, and the daily-refreshing free tier is the rare trial that lets you evaluate honestly for weeks. Our tier guidance: Apprentice for the unlocks, Artisan when volume or premium-model appetite outgrows it, Maestro only when queue priority becomes money.

The clear no: aesthetes chasing the single most beautiful image — Midjourney remains the stronger pure artist, and if your work is standalone visual impact over workflow, that's still the better subscription. Also mismatched: absolute beginners wanting one-button simplicity, who'll find Leonardo's model lists, sliders, and settings initially overwhelming (the learning curve pays off, but it exists — budget an afternoon with the community feed before judging); video-first creators, for whom Motion is a garnish and the dedicated video platforms are the meal; and anyone planning to build a business on free-tier output, which the ownership terms quietly discourage by design. Teams needing enterprise governance, indemnification, and procurement-grade contracts should also note that Leonardo — even under Canva — plays a tier below the Adobe/enterprise stack on legal wrapping, if a tier above it on creative flexibility.

The wider competitive map, for those still choosing: Midjourney owns pure aesthetics and an unmatched style culture, at the cost of weaker pipeline tooling and no free tier. Adobe Firefly owns legal safety and Creative Cloud integration, at the cost of house models that trail on flair — the enterprise's choice more than the artist's. Ideogram leads on typography-heavy design work like posters and logos. The open-source route — FLUX and Stable Diffusion weights self-hosted or through workbench tools — offers maximum control and zero subscription for the technically willing, and is effectively what Leonardo packages into a browser for everyone else; if you'd enjoy configuring that stack yourself, you're not Leonardo's customer, and if the previous clause sounded exhausting, you are. Canva's own generative features, fed increasingly by Leonardo technology, cover the casual end for people who just need a decent image inside a design. Leonardo's defensible ground against all of them is the same claim throughout this review: the strongest combination of control, consistency, and custom training per dollar — the production lane, held convincingly.

Practical playbook for subscribers: spend your first tokens learning which model fits your use case rather than defaulting to the newest — the community feed filtered to your genre is the fastest teacher. Train a custom model early if you have any recurring style or subject; it's the feature that compounds. Draft at standard settings and reserve enhancement passes for finals, since quality multipliers are where token budgets die. Turn on private generation the moment anything client-related enters a prompt. And export and archive systematically — your trained models and generation history are switching costs you should at least own copies of.

Weighing it: the best production toolset in consumer AI imaging, custom training and character consistency at unmatched price points, a genuinely sustainable free tier, and credible stewardship under Canva — against house models that trail the aesthetic frontier, a token system that demands literacy, free-tier IP terms that deserve a louder label, and a video feature that's a checkbox. That lands Leonardo a strong rating with a crisp identity: not the generator for the most beautiful image, but very possibly the generator for the next thousand images you actually need — consistent, controlled, and yours. The self-test is simple: if your last ten image needs were each one-offs, subscribe to the artist's tool; if they rhymed — same character, same style, same brand, same pipeline — Leonardo was built for exactly that rhyme, and the free tier will prove it to you within a week at zero cost.