You know the routine by now. You type a prompt into an AI image generator. “Fantasy book cover, enchanted forest, cloaked figure, dramatic lighting.” You hit generate. You wait. Four images appear. One is close but the figure is in the wrong spot. Another has the right mood but the composition is off. The third looks like it was painted by someone who has heard of forests but never actually seen one. The fourth is gorgeous, but it’s a portrait when you needed a landscape.
So you tweak the prompt. Regenerate. Wait. Four more almost-but-not-quite results. You’re twenty minutes and a dozen generations in, starting to suspect there must be a better way than feeding coins into a machine that keeps giving you the wrong snack.
JJ Fiasson thought so too.
The Philosopher Who Built a Creative Suite
Fiasson’s resume doesn’t follow a straight line. He studied philosophy and immunology at the University of Sydney (a combination that provokes more questions than it answers), earned a master’s in international security and international business, and spent twelve years running a cloud technology company before it was acquired by a publicly traded firm. He co-founded a fintech mortgage business that processed nearly a billion Australian dollars in loans. Then he started a gaming studio.
It was the gaming studio, Raini Studios, that brought Fiasson face-to-face with the problem he’d eventually try to solve. Creating visual assets for games was expensive and slow. Character designs took days. Environment art, texture work, and concept iterations were all bottlenecked by human production capacity and budget. When generative AI tools started emerging, Fiasson saw the potential immediately. He’d been watching the space since Google Deep Dream in 2015, waiting for the technology to catch up to the vision.
But the tools that arrived forced creators into an awkward choice. Midjourney offered beautiful output through a simple prompt box, but hid most of the creative controls. Running Stable Diffusion locally gave you every parameter imaginable, but required the kind of technical setup that sent non-programmers running. There was nothing in between. A tool with real creative depth that didn’t require a computer science degree to operate just didn’t exist yet.
Fiasson named his company after Da Vinci, who he described as “both an artist and an inventor.” The dual identity was the point. Leonardo.Ai launched in late 2022, and the growth was swift. By July 2024, under two years in, the platform had 19 million registered users. Canva acquired the company for a reported $320 million-plus, bringing all 120 employees into Canva’s team. Leonardo continues to operate as its own product while its AI capabilities get woven into Canva’s design platform.
For indie authors who already use Canva for cover design and marketing, that acquisition isn’t just corporate trivia. It means the gap between “generate an image” and “design a cover” is getting smaller.
More Than a Text Box
Leonardo.Ai is a full creative suite that happens to be powered by AI.
The core text-to-image generation works like you’d expect. Describe what you want, get an image. But where most generators give you a text box and a generate button, Leonardo gives you a cockpit. You can choose between multiple AI models (including their proprietary Phoenix model and third-party options like Flux). You can adjust guidance scale, seed values, and aspect ratios. You can upload reference images to guide style or character appearance.
For authors, the uses are familiar. Book cover concepts, character art, scene visualization, social media graphics. What’s different is how much control you get over each one.
Realtime Canvas lets you sketch rough shapes with a digital brush while the AI renders finished art around your strokes in near real-time. If you’ve ever struggled to describe a composition in words (“no, the figure should be on the LEFT, slightly smaller, with the castle BEHIND them”), you can just draw it. It feels less like prompting a machine and more like collaborating with an illustrator who can read your rough sketches.
Image Guidance offers separate controls for content reference (match the composition of an existing image), style reference (match the aesthetic), character reference (maintain a consistent character appearance), and you can mix them however you want. For series authors who need the same protagonist to look recognizably similar across five book covers, the character reference tools alone make Leonardo worth a serious look.
Phoenix, Leonardo’s in-house foundation model, tackles two persistent frustrations with AI image generation. First, prompt adherence. Phoenix follows long, detailed prompts more faithfully than many competitors, so when you describe a specific scene with multiple elements, more of them actually show up. Second, text rendering. Phoenix can produce coherent text within images (signs, book titles, labels) more reliably than average. It won’t replace adding your title in a design tool, but it’s a noticeable step up from the gibberish most generators produce. (You know the gibberish I’m talking about. We’ve all seen it.)
Universal Upscaler enlarges images to 8K resolution using generative refinement, which means it adds detail as it scales rather than just stretching pixels. If you’ve ever tried to use a standard-resolution AI image for a print book cover and watched it dissolve into a blurry mess at full size, you understand the value.
The Feature That Changes the Workflow
Flow State is the feature that made me reconsider how AI image generation should work.
Instead of the standard cycle (prompt, generate, evaluate four images, tweak, repeat), Flow State generates a continuous stream of images from a single prompt. They just keep appearing. You scroll through them like browsing a mood board that was built specifically for your description.
When something catches your eye, you click “More Like This,” and the stream redirects toward that aesthetic direction. Click again on a different image, and it narrows further. The process feels less like operating a machine and more like curating a gallery that updates in real time.
Each image in the stream costs a single token (Leonardo’s internal credit currency), making Flow State deliberately cheap. It’s exploration mode. Figure out what you actually want before committing real credits to a polished, high-resolution generation.
For authors, this solves a real problem. Describing exactly what you want for a book cover in a single text prompt is like trying to give driving directions using only cardinal directions. You can get there eventually, but it takes a lot of wrong turns. Flow State lets you browse your way toward the right image instead of trying to describe it perfectly upfront. You might discover a color palette or composition you never would have thought to request, simply because it scrolled past and something clicked.
It’s a small shift in how you interact with the tool. But if you’ve ever burned through thirty prompt revisions trying to land on a direction, it’s the kind of small shift that saves an evening.
What You Should Know Before Subscribing
Leonardo’s depth is also its biggest barrier. The interface is dense, with panels and sliders and model selection dropdowns spread across the screen. It can feel like walking into a recording studio when you just wanted to hum a tune. If you want the simplicity of typing a prompt and getting a beautiful image with minimal fuss, Midjourney’s cleaner interface is a more comfortable starting point. Leonardo rewards people who want control, and control comes with a learning curve.
Token math can get uncomfortable. Leonardo runs on a token-based credit system, and advanced features burn through tokens fast. The free plan provides 150 tokens per day, which sounds reasonable until you realize a single generation with the Alchemy v4 model can cost 8 to 16 tokens. Upscaling and video clips have their own token costs too, and the math isn’t always intuitive. Keep an eye on your balance until you develop a feel for what things cost.
The free plan is a test drive, not a tool. 150 daily tokens, limited resolution, and all your generations are public (anyone can see them). It’s enough to evaluate whether Leonardo fits your workflow. It’s not enough to produce real work. Plan on subscribing if you decide to use it seriously.
Video exists, but it’s a side dish. Leonardo can generate short video clips (up to 10 seconds), which is useful for social media teasers. But this is an image-first platform with video added on. For book trailers or longer promotional content, you’ll want a dedicated video tool.
Text rendering is improved, not solved. Phoenix handles text better than most competitors, but “better” still isn’t “reliable.” If legible text in your images is the primary need, Ideogram is the specialist for that job. For Leonardo, plan to add titles and author names in Canva or another design tool.
Everything lives in the cloud. No offline mode, no local processing. If Leonardo’s servers go down or you’re working somewhere without reliable internet, your tools go with them.
Who This Is For
Leonardo.Ai fits best if you actually enjoy having creative control over your visual assets and you’re willing to invest time learning the interface. If you’re the kind of person who likes adjusting settings and fine-tuning until you get exactly what you envisioned, Leonardo gives you more to work with than almost any other consumer-facing image generator.
Series authors, in particular, should pay attention. The character and style reference tools, combined with the ability to train custom models on your own reference images, make it easier to maintain visual consistency across multiple book covers than most competitors manage.
And if you already use Canva for your cover layouts and marketing materials, Leonardo’s growing integration with Canva’s platform means you’re getting closer to a single pipeline: generate the image, design the cover, export for publishing.
Who This Is Not For
If your ideal workflow is “type a prompt, get something gorgeous, move on,” Midjourney is a better match. Leonardo rewards tinkering, and not everyone wants to tinker. That’s totally valid.
If you only need images occasionally, like one cover concept per book and maybe some social media graphics around launch, the subscription is hard to justify. Paying $12 to $30 per month makes sense for regular use, not for a handful of images a year.
And if you need perfect text rendered directly into your images without post-processing, Ideogram remains the specialist for that.
If nothing else, try the free plan and spend ten minutes in Flow State. It changes the creative process from “describe what you want perfectly in words” to “recognize what you want when you see it.” For anyone who’s burned an entire evening in the prompt-tweak-regenerate loop, that alone might sell you on it.
