Leonardo AI

It started with a problem every game developer knows intimately: creating visual assets is expensive, slow, and creatively unforgiving. One wrong character design, and you're back to the drawing board — literally. JJ Fiasson and his four co-founders started Leonardo AI in Sydney in 2022 not to build another AI art toy, but to build a production pipeline. Something that could generate a thousand consistent game assets, train on your specific art style, and give a solo indie developer the visual output of a full studio art team.

That focus on production consistency — not just impressive one-off outputs — is what set Leonardo apart in a crowded field. By early 2025, the platform had crossed 30 million users. In July 2024, Canva acquired the company, adding Leonardo's model infrastructure and generation capabilities to Canva's 260 million monthly users. It is now the AI backbone powering Canva's Magic Media feature and bundled directly into the new Canva Business plan.

Phoenix: The Flagship Model

Leonardo's proprietary Phoenix model is the creative heart of the platform — and the thing most competitors quietly benchmark against. Phoenix 1.0 generates photorealistic humans, complex multi-element compositions, and detailed textures with a level of anatomical accuracy and lighting coherence that earlier diffusion models consistently failed at. Its successor, Phoenix 2.0, adds real-time generation via Flow State: a mode so fast and fluid that it feels less like prompting an AI and more like sketching with one.

"We call it Flow State because it's so fast and seamless — it encourages creative exploration. It can do anything from a cute creature to photorealism to images with text in them." — JJ Fiasson, CEO

Not Just One Model — A Full Creative Stack

Since the Canva acquisition, Leonardo has expanded far beyond its own models. The platform now integrates:

• Google Veo 3 — cinematic video with real audio and environmental sound generation
• Kling 2.5 — high-fidelity video with precise motion control
• Motion 2.0 and 2.0 Fast — Leonardo's own video generation models for fluid, consistent animation
• ElevenLabs audio — voice and sound generation directly inside the Maestro tier
• Topaz upscaling — export-ready 4K asset enhancement

The result is one of the most complete generative media stacks available at consumer price points — image, video, and audio in a single subscription, no tab-switching required.

Custom Model Training: The Power Feature

Leonardo's most differentiated capability — the one that built its reputation in the game development community — is custom model training. Upload 20–50 images of a character, environment, object, or art style, and Leonardo trains a fine-tuned model on your dataset. Every subsequent generation inherits that style with uncanny consistency. For game studios and brand teams managing visual identity at scale, this is not a convenience feature. It's the entire value proposition.

The API for Builders

Leonardo's production API serves thousands of applications — design tools, marketing platforms, e-commerce product photography pipelines, and game engines that need generative assets on demand. The platform recently transitioned to Pay-As-You-Go pricing for new API users, removing the friction of subscription minimums for developers who want to experiment before committing. A visual pricing calculator lets builders estimate costs per generation before writing a line of code.

30 Million Users. $31M Raised. Acquired for an Undisclosed Sum.

Backed by Blackbird Ventures, Smash Capital, and TIRTA before its acquisition, Leonardo raised $31M in two rounds and grew to 30M+ users without the kind of marketing spend its US-based competitors deployed. Canva did not disclose the acquisition price, but given Leonardo's user scale and the strategic value of its model infrastructure to Canva's AI roadmap, industry estimates placed the deal in the $100M–$200M range.