✓ Reviewed by The Future Signal
✓ Reviewed by The Future Signal
Leonardo AI offers more creative control than almost any competitor, with custom LoRA model training, character consistency and a real-time editing canvas that make it the strongest choice for repeatable asset production. Its free tier is generous at 150 tokens daily but unsuitable for commercial work, since generations are public and Leonardo holds the IP. Token pricing is hard to forecast and video is poor value. Owned by Canva since 2024, with genuine roadmap uncertainty ahead.
Leonardo’s business impact concentrates in asset production rather than one-off creative work. Custom LoRA model training lets brands build a model on their own products or visual style, producing on-brand output at volume without repeated art direction. Character and style reference solve the consistency problem that makes most AI image tools impractical for campaign sets, storyboards or game assets.
For e-commerce teams, generating product and lifestyle imagery in-house replaces photography retainers at meaningfully lower cost. The offsetting consideration is commercial terms: free-tier output cannot be used commercially, and Leonardo offers no IP indemnification, so legal exposure sits with the customer.
Compare plans and pricing to find the best option for your needs.
Compare plans and pricing to find the best option for your needs.
Last Updated: 28 August 2026
Leonardo AI gives creators more control than almost any competitor — custom model training, a real-time canvas, consistent characters, and a model picker spanning its own Phoenix and Lucid Origin alongside FLUX, Veo and others. Its free tier is unusually generous at 150 tokens a day. It is also the free tier businesses should read carefully, because free generations are public and, according to Leonardo’s own pricing page, the intellectual property sits with Leonardo rather than you. Owned by Canva since 2024, the product remains independent but the long-term roadmap is genuinely uncertain.
Estimated reading time: 10 minutes
Leonardo AI is a Sydney-based generative platform for images and video, founded in December 2022 by JJ Fiasson and five co-founders. It started as a tool for generating game art and grew into a broader creative suite, reaching more than 30 million users and over two billion images generated by early 2026.
Its distinguishing characteristic has always been control. Where Midjourney optimises for beautiful results with minimal input, Leonardo gives you levers: custom model training, character reference, a real-time editing canvas, negative prompts, and fine-grained generation settings. That made it popular with game designers, illustrators and asset production teams rather than casual users.
The platform runs both in-house and third-party models. Its own include Phoenix — described as Australia’s first generative AI foundation model — and Lucid Origin. Alongside these sit partner options including FLUX, Veo 3 and Kling, so users can pick the engine that fits a given job.
Canva acquired Leonardo in July 2024 in a deal reported at more than $320 million, bringing its entire 120-person team across. Leonardo continues to operate as an independent platform with its own interface and subscriptions, while Phoenix now powers parts of Canva’s own AI features including Dream Lab.
Why this matters for your business: Leonardo is a production tool, not a novelty generator. Its strengths — repeatable characters, trained brand models, batch asset work — are the things that matter when you need forty consistent images rather than one impressive one. But its terms and token economics need reading properly before you build a workflow on it.
Leonardo AI is a strong fit for:
Leonardo AI is probably not the right choice for:
Phoenix and Lucid Origin. Leonardo’s in-house foundation models, built for prompt adherence and image quality. These are the models covered by unlimited relaxed generation on paid plans.
Third-party model access. FLUX, Veo 3, Kling and others available in the same interface, letting you match model to task without changing platform.
Custom model training (LoRA). Train a model on your own reference images to lock a style, character or product appearance. Apprentice includes one training slot monthly, Artisan around twenty personal model slots, Maestro around fifty.
Real-time canvas. Sketch and edit while generation happens, giving directed control rather than prompt-and-hope. One of the clearest differentiators from prompt-only competitors.
Character and style reference. Keep a character or aesthetic consistent across many generations — the feature that makes Leonardo viable for storyboards, game assets and campaign sets.
Universal upscaling and background removal. Practical production tools including transparent PNG output, which matters more for real asset work than another style preset.
Relaxed mode. When fast tokens run out, Artisan and Maestro users can generate without token cost in a slower queue. Important caveat: this covers basic models only, not premium fine-tuned or third-party ones.
API access. Available with a starter credit and no expiry, plus a pricing calculator for estimating consumption before committing.
FREE PLAN ──────► 150 tokens/day, no rollover
│ ⚠ public generations
│ ⚠ IP held by Leonardo
▼
PAID PLANS ─────► private images, you hold IP
│
├─ Leonardo's own models ──► unlimited in relaxed mode
│ (Artisan/Maestro only)
│
├─ Premium third-party ────► ALWAYS costs tokens
│ FLUX · Veo · Kling no relaxed fallback
│
└─ Video ──────────────────► ~2,500 tokens per 5s clip
≈ 10 clips on a $30 plan[Illustration placeholder: Leonardo AI token consumption map]
Leonardo sits between Midjourney’s simplicity and self-hosted Stable Diffusion’s complexity, and that is a deliberate position. The interface is a normal web app rather than a Discord bot, models are chosen from a dropdown, and most people generate something usable within minutes.
The learning curve is the token system. Costs vary by model, resolution and quality setting, ranging from a couple of tokens for a basic image to thousands for a video clip. Nothing in the interface makes that intuitive in advance, and users report frustration when failed generations still consume credits.
The other thing that takes learning is what “unlimited” means. Paid marketing emphasises unlimited relaxed generation, but this applies only to Leonardo’s own basic models. If your workflow depends on FLUX or Veo, you are on metered tokens regardless of tier.
Future Signal Tip: Before subscribing, spend a week on the free plan generating exactly what you actually need, and track token consumption per image. Then multiply. Most people either overpay for Maestro or run dry on Apprentice because they estimated from image counts rather than token costs.
Leonardo’s image quality is good and, more importantly for production work, controllable. Phoenix and Lucid Origin handle prompt adherence well, and the character reference and style tools deliver the consistency that asset work requires. For generating forty variations that look like they belong together, Leonardo is stronger than most competitors.
It does not lead on pure aesthetics. Independent commentary generally places Midjourney ahead for showcase imagery and FLUX ahead for photographic realism — which is precisely why Leonardo hosts FLUX as an option rather than pretending otherwise.
Video is the weakest value proposition. Generating a five-second Veo 3 clip reportedly consumes 2,500 fast tokens, which is roughly ten clips before exhausting a $30 Artisan plan. Dedicated video platforms offer better economics and more video-specific controls. Leonardo’s video is a convenience for image-first teams, not a reason to choose the platform.
| Use case | How Leonardo performs | Better alternative |
|---|---|---|
| Consistent characters and assets | Excellent — the core strength | — |
| Custom trained brand models | Excellent, rare among competitors | — |
| Directed editing and iteration | Strong, via real-time canvas | — |
| Pure aesthetic quality | Good, not leading | Midjourney |
| Photographic realism | Good via FLUX partner model | FLUX directly |
| Video generation | Expensive per clip | Runway, dedicated tools |
| Commercially indemnified output | Not offered | Adobe Firefly |
This is the part businesses most need to read, and it is easy to miss.
Leonardo’s free plan is genuinely generous on volume: 150 fast tokens daily, enough for roughly 25 to 50 images depending on settings, with no rollover. That is more than most competitors give away, and it makes evaluation easy.
The terms are different from the volume. According to Leonardo’s pricing page, free-tier creations are public — visible to and usable by other Leonardo users — and you do not hold the intellectual property in them. Leonardo does.
For a hobbyist experimenting, that is a fair trade. For a business generating product visuals, brand assets or client work, it is disqualifying. Paid plans resolve both issues, giving private generations and IP ownership, which is the real reason to upgrade rather than any token allowance.
⚠️ The practical rule. Use the free plan to evaluate the models and learn the token system. Do not use it for anything you intend to publish, sell, or show a client. The upgrade to Apprentice at $12 a month is inexpensive precisely because it is not optional for commercial work.
Leonardo’s integration story is narrower than platform competitors. There is an API with a starter credit and a pricing calculator, aimed at teams building generation into their own pipelines, and Canva ownership means some Leonardo capability now appears inside Canva’s own products.
What is absent is a broad connector ecosystem. No CRM integrations, no automation platform connectors, no native presence inside design suites the way Firefly sits in Photoshop. Leonardo is a destination you go to, not a layer across tools you already use.
One integration point deserves flagging. Leonardo’s model list has included Sora 2 among its third-party options, and OpenAI has scheduled the Sora API for shutdown on 24 September 2026. Any workflow built around that specific model should be reviewed, and it is a useful reminder that partner model availability on any platform depends on decisions the platform does not control.
Leonardo charges in tokens rather than per image, which makes it cheap to start and easy to misjudge at scale.
| Plan | Monthly | Annual (per month) | Tokens |
|---|---|---|---|
| Free | $0 | — | 150/day, no rollover |
| Apprentice | $12 | ~$10 | ~8,500/month |
| Artisan | $30 | ~$24 | ~25,000/month |
| Maestro | $60 | ~$48 | ~60,000/month |
| Teams | From ~$24/seat | — | Multi-seat, three-seat minimum |
Annual billing saves roughly 20% and reported figures vary slightly by source, so confirm current rates directly before committing.
Three things determine whether this is good value. First, whether your work is commercial — if so, the free tier is off the table and $12 is your genuine floor. Second, whether you use Leonardo’s own models or premium third-party ones, since only the former get unlimited relaxed generation. Third, whether you generate video, which consumes tokens at a rate that makes Leonardo poor value against dedicated platforms.
For image-first asset production on Leonardo’s own models, Artisan at $30 is the sweet spot: custom model training, real-time canvas, and unlimited relaxed fallback when fast tokens run out. For video-led work, look elsewhere.
Reviewers also report recurring complaints about refunds and failed generations consuming credits, which marketing pages do not mention. Worth knowing before you commit annually.
How many images do you actually need per month, and on which models? That question answers your plan tier better than any feature comparison.
Midjourney — better pure aesthetics with far less control. The right choice when you want beautiful results rather than repeatable ones.
Adobe Firefly — the option when commercial legal cover matters, offering IP indemnification on its own models that Leonardo does not match.
Canva — worth noting that Leonardo’s Phoenix model already powers parts of Canva’s AI features, so Canva users may have some of this capability already.
Runway or dedicated video tools — meaningfully better economics and controls if video is your primary output.
FLUX directly — cheaper if you will mostly use FLUX anyway, at the cost of Leonardo’s workflow tools.
Leonardo’s position in 2026 is a study in how acquisitions actually play out. Canva bought it in 2024, kept the team, kept the brand, and let the product run independently — while quietly folding Phoenix into its own Dream Lab and Magic Studio features.
That has been good for users so far. Pricing has not shifted materially, the team stayed, and the product kept shipping. But the strategic logic points one way: Canva bought a foundation model and a research team, not a standalone subscription business. Leonardo’s roughly 30 million users are a rounding error against Canva’s hundreds of millions.
The plausible futures are tighter integration, continued independence, or eventual absorption. Nobody outside Canva knows which, and the honest advice for heavy users is to keep a backup workflow rather than assume permanence.
The wider signal is about model dependency. Leonardo hosts third-party models whose availability it does not control — the Sora retirement is a live example. Any platform built on other people’s models inherits their decisions, and that is worth factoring into a multi-year plan.
Overall Rating: 7.7 / 10
Leonardo AI is the best choice available for controlled, repeatable asset production. Custom model training, character consistency and the real-time canvas give creators levers that Midjourney does not offer, and the free tier makes evaluation genuinely easy.
It loses points for token pricing that is hard to estimate, video economics that do not compete, an absence of commercial legal protections, and roadmap uncertainty under Canva. Most importantly, the free tier’s public-generation and IP terms make it unsuitable for any commercial use.
Future Signal recommends Leonardo AI for game developers, e-commerce teams and brand asset production, on a paid plan. Evaluate free, publish paid, and keep a backup workflow given the ownership question.
Can I use Leonardo AI images commercially? Only on a paid plan. Leonardo’s pricing page states that free-tier creations are public and that Leonardo, not you, holds the intellectual property. Paid plans provide private generations and IP ownership, which is the main reason to upgrade.
How much does Leonardo AI cost? Free with 150 tokens daily, then Apprentice at $12 a month, Artisan at $30 and Maestro at $60, dropping to roughly $10, $24 and $48 on annual billing. Teams start around $24 per seat with a three-seat minimum. Reported figures vary slightly, so confirm directly.
What does “unlimited generations” actually include? Unlimited relaxed-mode generation on Leonardo’s own basic models, available on Artisan and Maestro. Premium third-party models including FLUX, Veo and Kling always consume tokens, with no relaxed fallback.
Is Leonardo good for video? It works, but the economics are poor. A five-second Veo 3 clip reportedly costs 2,500 fast tokens — around ten clips on a $30 plan. Dedicated video platforms offer better value and more video-specific controls.
Does Canva owning Leonardo affect me? Not so far. Leonardo retained its team, brand and independent product, and pricing has not shifted materially. The long-term direction is genuinely uncertain, so heavy users should keep an alternative workflow ready.
Leonardo or Midjourney? Midjourney for aesthetic showcase work with minimal input. Leonardo for controlled, repeatable asset production — custom trained models, consistent characters, directed editing. Leonardo also has a free tier; Midjourney does not.
What is LoRA training and do I need it? LoRA training builds a custom model on your own reference images to lock a style, character or product look. It is Leonardo’s clearest differentiator and genuinely valuable for brand consistency. Apprentice includes one slot monthly, with more on higher tiers.
Does Leonardo offer IP indemnification? No. Unlike Adobe Firefly, Leonardo does not offer indemnification or a licensed-training-data guarantee. Businesses needing legal cover for AI output should factor that into their choice.
Leonardo AI earns its place by giving creators control that most competitors deliberately hide. Custom trained models, consistent characters, a real-time canvas and a genuine model picker make it the strongest option available for repeatable asset production — the kind of work where you need forty images that belong together rather than one that impresses.
Its weaknesses are practical rather than fundamental. Token pricing is hard to forecast, video is poor value, and there is no legal protection comparable to Adobe’s. The ownership question under Canva adds uncertainty that heavy users should plan around rather than ignore.
The one thing worth acting on immediately is the free tier. It is generous, useful and completely unsuitable for commercial work, because your generations are public and the IP is not yours. Evaluate freely, publish on a paid plan, and track your token consumption before choosing a tier. Do that and Leonardo is a genuinely strong production tool at a fair price.
Our final assessment after evaluating features, performance, value, and business impact:
Leonardo AI earns a recommendation for game developers, e-commerce operators and brand teams producing visual assets at volume, where its custom model training, character consistency and real-time canvas deliver control that Midjourney deliberately withholds. The free tier makes evaluation genuinely easy.
Two cautions matter. First, free-tier generations are public and Leonardo holds the intellectual property, so no commercial work should ever run on it — the $12 Apprentice plan is a floor, not an option. Second, token pricing is difficult to forecast, video economics are poor against dedicated platforms, and there is no IP indemnification comparable to Adobe Firefly. Canva ownership adds roadmap uncertainty worth planning around.
We may earn a commission if you purchase through our links, at no extra cost to you. This helps support independent reviews.
Explore other highly rated AI tools we’ve reviewed in this category.
