Key Features
- Phoenix model (proprietary, best-in-class prompt adherence)
- 150 free daily tokens (resets every 24h)
- Text-to-image generation with 20+ models
- AI Canvas (inpainting, outpainting, layered editing)
- Realtime Canvas (draw a sketch → AI generates in real time)
- Motion (animate still images into short video clips)
- Text-to-video via Veo 3, Kling 2.5, Sora 2, LTX-2
- 3D Texture Generation (.OBJ files for game engines)
- Custom model training (upload 15–30 images to train your own style)
- Character reference models (consistency across poses)
- Universal Upscaler (up to 4x, no quality loss)
- Transparent PNG Maker (background removal)
- Prompt expansion via Gemini integration
- Image-to-image generation
- Elements (LoRA-based style modifiers)
- API access (separate plans, credits never expire)
- Private mode (paid plans only)
- Unlimited Relaxed Generation (Artisan and above)
What Is Leonardo AI?
Leonardo AI started with a specific problem in mind: game developers needed fast, controllable visual assets without paying for a stock library or a full design team. That origin shows in the product. The platform has granular controls, a canvas editor with real inpainting and outpainting, custom model training on your own image sets, and 3D texture generation for .OBJ files — features that read like a game dev's wish list and ended up attracting everyone else in the process.
The core workflow is text-to-image, but the models matter here. Leonardo gives you access to its own Phoenix model, plus a long list of third-party models including Flux variants, Ideogram 3.0 for typography, Seedream for high-resolution cinematic output, and video models like Veo 3, Kling 2.5, and Sora 2. The tradeoff is that each model has different token costs and different use cases — which means there's a real learning curve before you know which tool to reach for. The Realtime Canvas is genuinely different from what any other major platform offers: you sketch rough shapes on the left, and the AI renders the result on the right in real time, updating as you draw. It's faster for ideation than typing and revising prompts.
The Canva acquisition in July 2024 changed the trajectory. Canva built a $200 million fund to compensate creators whose work is used in AI training — positioning the combined platform toward the ethical end of an industry still navigating significant legal gray zones. Whether that matters to your use case is your call, but for enterprises with legal teams paying attention to AI provenance, it's a meaningful signal.
Where Leonardo can frustrate: the token system rewards heavy users and punishes inconsistent ones. Unused monthly tokens don't carry over. Video generation costs are steep. Third-party models charge tokens even if you're on a paid plan. And the free plan's public gallery means everything you generate at no cost is visible to other users — not a dealbreaker, but something to be aware of before you start generating client work on the free tier.
Best for
Game DevelopersDigital ArtistsConcept ArtistsMarketing TeamsContent CreatorsE-commerce BrandsSocial Media ManagersDesignersDevelopers (API)Educators
Use cases
Game asset creation (characters, environments, textures, UI)Marketing visuals and ad creativesSocial media graphics and short animationsProduct mockups and e-commerce imageryConcept art and storyboardingYouTube thumbnails and channel artBrand visual development with custom model training3D texture generation for game enginesShort video clips and animated loopsBlog and editorial illustration
Key features explained
Phoenix Model — Prompt Adherence Done Right
Phoenix is Leonardo's first proprietary foundational model, built from scratch rather than adapted from an existing architecture. Its defining trait is prompt adherence — it follows detailed, multi-part instructions more faithfully than most competing models. Ask for a poster with the word "SALE" in bold red letters over a blue background with a specific arrangement, and Phoenix delivers the text correctly spelled and positioned. That sounds basic, but it's been a consistent failure point for AI image generators since the beginning. Phoenix also handles restyling well — feed it a rough sketch or 3D block-out and it renders a finished asset while preserving the original structure. For creative directors, concept artists, and marketers who need the output to match the brief, Phoenix is the starting point. Run it in Quality mode rather than Fast for best results; Fast mode saves tokens but cuts coherence noticeably on complex prompts.
AI Canvas and Realtime Canvas — Real Editing, Not Just Generation
Most AI image tools give you a generation and a download button. Leonardo gives you a canvas. The AI Canvas has layers, masks, brushes — and two features that save more time than anything else in the platform. Inpainting: you circle a broken area (hands are the classic example — AI has always struggled with hands), type what it should look like, and the model fixes just that spot without disturbing anything around it. The lighting matches. The texture matches. It's not perfect every time, but it's faster than regenerating the whole image and hoping. Outpainting works the opposite way — extend the borders of any image and let the AI fill in what's beyond the frame. A portrait becomes a full-body shot. A vertical image becomes a horizontal banner. Then there's the Realtime Canvas, which is a genuinely different experience: rough shapes on the left, rendered result on the right, updating live as you draw. Faster for ideation than typing a prompt, waiting, reading the result, adjusting the prompt, repeating.
Custom Model Training — Your Style, Reproducible
This is the feature that changes how teams work, and most people discover it by accident. Upload 15 to 30 images — a character design, a product line, a brand photography style — and Leonardo trains a model specifically on that dataset. From that point, every generation using that model pulls from your reference material. Same face across different poses. Same product style across different backgrounds. Same illustration aesthetic without writing a paragraph of style instructions every single time. A marketing team can train on product photos and generate on-brand visuals without a designer in the loop. A game studio can build a character sheet, train on it, and iterate through environments and expressions with consistent results. It's available on all plans including free — which, given that competitors charge for this at higher tiers, is a genuine differentiator. Elements (Leonardo's LoRA-based style modifiers) sit below this on the complexity scale: lighter, faster, no training required, applied on top of any model as an overlay. Good for quick aesthetic experiments. Custom training is for when you need results you can actually rely on.
Video Generation — Powerful but Token-Heavy
Leonardo integrates a full suite of video models: Veo 3 and Veo 3.1, Kling 2.5 Turbo and Kling 2.1 Pro, Sora 2 and Sora 2 Pro, LTX-2 Pro and LTX-2 Fast, plus Leonardo's own Motion 1.0 and 2.0 for animating still images. The quality gap between models is significant — Veo 3 produces the most cinematic results, while Motion 1.0 handles simple parallax animation at lower token cost. The critical thing to understand before you start generating video: token costs are steep. A single 8-second Veo 3 clip uses roughly 300 tokens — equivalent to an entire day of free-plan image generations. Motion and LTX-2 Fast are considerably cheaper and eligible for Relaxed Generation on Artisan and above. Third-party video models are never included in Relaxed Generation regardless of plan. If video is a primary use case, run the numbers on your expected monthly volume before choosing a plan — the math is different from image-focused workflows.
Pricing Plans — Free, Apprentice, Artisan, Maestro
The Free plan gives you 150 tokens daily, resetting every 24 hours. No credit card required, commercial license included (non-exclusive), but all images are public. Apprentice at $12/month ($10 on annual billing) adds private mode, priority generation speed, and 8,500 monthly tokens — the entry point for anyone generating client work. Artisan at $30/month ($24 annual) steps up to 25,000 monthly tokens plus Unlimited Relaxed Generation for first-party models, meaning you can keep creating images after your token balance hits zero, just at lower priority. Maestro at $60/month ($48 annual) provides 60,000 monthly tokens and up to 6 concurrent generation jobs — suited for agencies and high-volume workflows. API access is sold separately, starting at $9/month with credits that never expire — a meaningful distinction from subscription tokens. Key thing to plan around: paid plan tokens do not roll over between billing cycles. If your production volume is irregular, monthly billing may give you more flexibility than locking into annual pricing.
Pros & Cons
Pros
- Most generous free tier of any major AI image generator — 150 tokens daily with no credit card required
- Phoenix model consistently outperforms competitors on complex, text-heavy prompts
- AI Canvas and Realtime Canvas offer Photoshop-level editing without leaving the browser
- Custom model training is available on all plans including free — rare in this category
- Acquired by Canva in 2024, which means stability, resources, and growing integration with a platform used by 190M+ people
- No Discord required — clean web interface accessible to anyone
Cons
- Video generation burns tokens extremely fast — a single 8-second Veo 3 clip costs as much as ~300 basic images
- Free plan images are public by default — anyone can view and use them from the community gallery
- Token system is confusing for new users — advanced models cost more tokens than expected, and unused tokens don't roll over
- Many of the standout third-party models (Nano Banana, Veo 3, Kling) charge tokens even on paid plans, and are often cheaper directly elsewhere
- Inconsistent quality across community-trained models — requires trial and error to find reliable ones
- Character consistency is still imperfect — the same character can look noticeably different across multiple generations
Frequently Asked Questions
Is Leonardo AI free to use?
Short answer: yes. Longer answer: it depends on what you're making. The free plan gives you 150 tokens a day, reset every morning — no credit card, no trial countdown. On a basic model that's anywhere from 20 to 75 images, which is a lot of experimenting. The catch nobody tells you upfront is that your images land in the public community gallery. If you're playing around, fine. If you're mocking up something for a client and don't want it visible to two million strangers, you'll need to upgrade. The other thing worth knowing: free plan commercial rights exist but they're non-exclusive — Leonardo keeps rights to the images too. For personal projects that doesn't matter. For anything going into paid advertising or client deliverables, a paid plan is the safer call.
Leonardo AI vs Midjourney — which is better?
Depends on the job. Midjourney still has an edge on raw aesthetic output — the images look expensive, the color work is strong, and most professional illustrators I know use it as their default. But Midjourney runs through Discord, has no free plan, and gives you almost no editing tools. You generate, you download, you're done. Leonardo is messier in some ways — more models, more settings, more decisions — but it gives you an actual canvas editor, inpainting, outpainting, custom model training, and a free tier that resets daily. Phoenix beats Midjourney significantly on text rendering inside images: logos, signs, labels with specific wording. For game developers, marketers, and teams producing high volumes of iterative work, Leonardo's toolset makes more practical sense. For illustrators who want one beautiful image and don't need to edit it afterward, Midjourney is still the benchmark. They're not really competing for the same user.
How does the token system work?
Tokens are Leonardo's usage currency. Almost every action on the platform costs tokens — generating an image, running the upscaler, creating a video, using the canvas editor. The cost varies significantly by model and quality setting: a basic image on a standard model runs 2–5 tokens, while the same image in Ultra Mode on a premium model can cost 20–30 tokens. Video is where things get expensive fast — a single 8-second Veo 3 generation costs around 300 tokens at standard quality, the equivalent of an entire day's free allowance. Free plan tokens reset daily and don't accumulate. Paid plan tokens are monthly and don't roll over — if you generate 5,000 tokens worth of content in week one and go quiet for three weeks, you lose the remainder at the end of the billing cycle. Artisan and Maestro plans include unlimited Relaxed Generation, which lets you keep creating after tokens run out — just at lower priority and slower speeds.
