Leonardo AI

Where it sits — and who owns it

Leonardo AI is a Canva company. That fact needs to lead any honest review in 2026. Founded in Sydney in December 2022 by JJ Fiasson and co-founders, Leonardo grew into one of the most popular image generation platforms outside the big US labs — third-party trackers put it past 30 million registered users and 2 billion images generated — and Canva announced its acquisition on July 29, 2024. Financial terms were never officially disclosed; reporting has variously placed the deal around US$200 million and north of US$320 million, with a restructure report in early 2026 citing US$370 million. The deal closed around mid-2025. Then came the harder chapter: in early 2026 Canva restructured Leonardo's video unit, folding technical staff into Canva's core AI operations, with reporting describing the move as effectively ending Leonardo's status as an independent entity. Canva's public position is that it "will continue to invest in evolving Leonardo as a stand-alone product" with an expanded feature set.

Anyone who used Leonardo through 2023 and 2024 will remember the pace: new model families, Alchemy, PhotoReal, canvas tools and video landing in short succession. That cadence is the through-line now under Canva. So the product continues, actively — but the ownership question changes what a buyer is signing up for. Leonardo is now both a product and an R&D feed for Canva's Magic Studio, and its Phoenix model already powers parts of Canva's own AI features. For individual subscribers this is mostly good news (backing, models, longevity). For teams betting a pipeline on Leonardo's API, it is a factor to weigh alongside the roadmap.

What it actually does

Leonardo began as a game-asset generator with its own fine-tuned models and has repositioned as a multi-model creative studio: one web interface, one token wallet, and a rotating catalog that mixes Leonardo's own foundation models (Phoenix, Lucid Origin) with licensed frontier models — FLUX, Google's Veo, Kuaishou's Kling, Ideogram, GPT Image, Nano Banana, Seedream and more (the API docs list 50+ model guides as of October 2026). The workspace includes image generation with style and element controls, canvas-based editing and inpainting, an upscaler, a real-time sketch-to-image canvas, video generation (Image-to-Video / Motion), consistent-character tooling via Character Reference, and LoRA training for custom styles. Team plans add shared asset libraries and model collections. The model catalog has become Leonardo's strategic center of gravity: where the product once sold its own fine-tuned models as the reason to subscribe, it now sells access — the ability to swap between a dozen frontier models without a dozen accounts, each priced in the same tokens. For many buyers in 2026 that convenience, not raw model quality, is the purchase argument.

Pricing at time of writing: a free tier with 150 tokens per day, resetting every 24 hours, on public generations; Apprentice at US$12/month (US$10 billed annually) with 8,500 tokens per month, private generations and one LoRA training per month; Artisan at US$30/month (US$24 annually) with 25,000 tokens, premium models and the upscaler; Maestro at US$60/month (US$48 annually) with 60,000 tokens and the highest concurrency. The Production API is separate — pay-as-you-go in USD with a top-up balance and a pricing calculator, and app subscriptions do not include API access. Token costs per image vary by model and feature: community pricing analyses put a standard polished image around 10 tokens, with Alchemy, PhotoReal, upscaling and video costing more.

My testing notes

Two observations from my testing. First, token economics require attention. My free-tier budget — 150 tokens — disappeared inside a single focused session once I used the premium pipelines; roughly a dozen considered generations with refinement, versus a couple of dozen quick drafts on the basic models. On paid plans the math gets friendlier (Artisan's 25,000 tokens at ~10 tokens per polished image is a few thousand images a month), but heavy video use will still hit Maestro-level budgets fast. The daily-reset free tier remains one of the more honest free offerings in the category for evaluation purposes.

Second, the aggregator shift is real and it changed how I use the tool. I ran one product-visual prompt through Phoenix, a FLUX variant and a Google model in the same session, on the same token wallet, with the same reference image — and the differences between models became a dial to tune rather than a reason to switch subscriptions. Phoenix remains the sweet spot for stylized game and product art, while the licensed models handle photorealism and text-heavy layouts better than Leonardo's own. Character Reference held face identity across outfit and background changes well in my runs — better than reference-image workflows in Firefly, short of dedicated tools like Scenario's trained models — which is the feature game and comic teams buy it for. Third, the upscaler and canvas tools are genuinely production-adjacent: I took a rough generation through inpaint-into-upscale to a 2K deliverable without leaving the app, and only reached for external editing when I needed typography, which remains Leonardo's weakest point.

How it compares in practice

Against Midjourney, Leonardo trades peak aesthetics for a free tier, editing tools and model choice. Against Firefly, it trades corporate provenance story for a far wider model catalog and game-asset heritage. Against Scenario, it trades trained-style consistency at scale for accessibility and price. Those are the real decision axes in this category now, and Leonardo sits deliberately in the middle of all three.

Where it falls short

The token system is Leonardo's standing complaint: costs differ per model and per feature, the free tier caps at a daily 150 tokens, and private generations on Apprentice are themselves limited per month. The platform's breadth is also its maintenance burden — features arrive fast and sometimes feel half-integrated, particularly in video. API access is pay-as-you-go on a separate wallet from app subscriptions, so the "I'll just script it" buyer faces a second pricing relationship. Then the ownership tail risk: the 2026 restructure moved engineering talent toward Canva's core, and while the product roadmap continues, feature priority now sits inside a much larger company's plans. For enterprise buyers there is also the data question: Leonardo's terms and Canva's broader data practices should be read together now, and the restructure reporting noted the unit was unprofitable with heavy cash burn before acquisition, which historically predicts pricing pressure on the consumer tiers. And absolute quality per image still trails Midjourney on aesthetics and FLUX on prompt adherence in my view; Leonardo's edge is workflow and consistency, not peak output.

Verdict

Rated 3.9 on the ToolSpotter rubric: core functionality depth is broad — multi-model generation, canvas editing, video, character consistency and custom training (the 40%); value is decent with a genuinely usable free tier and US$12 entry, though token opacity and the separate API wallet dent it (the 20%); usability is good for creators who learn the token system, mediocre for first-timers (the 20%); the model catalog and Canva ecosystem are strong (the 10%); and reliability signals are the open question — huge user scale and Canva's backing on one side, the 2026 restructuring and undisclosed roadmap control on the other (the 10%). For game and marketing creatives who want consistency and a model buffet, Leonardo remains a strong buy; for pure image quality, look at Midjourney or FLUX-based tools — and re-check the pricing page, because token allotments have moved twice in the last year.

Sources

• https://canva.com/newsroom/news/leonardo-ai
• https://ainvasion.com/leonardo-ai
• https://javisreview.com/review/leonardo-ai-review-2026
• https://bbrw.com/canva-restructures-leonardo-ai-ai-video-unit-18-months-after-370-million-acquisition
• https://neuronfeed.com/startups/leonardo-ai
• https://pricemyai.com/pricing/leonardo
• https://techsifted.com/guides/leonardo-ai-pricing-2026
• https://dynalord.com/blog/leonardo-ai-pricing
• https://apis.io/plans/leonardo-ai/leonardo-ai-plans-pricing
• https://costbench.com/software/ai-image-generators/leonardo-ai
• https://github.com/api-evangelist/leonardo-ai/blob/main/plans/leonardo-ai-plans-pricing.yml