Leonardo.ai is an AI image platform built for creative professionals who need more control than a simple text box. Custom model fine-tuning, a library of community-trained style models, character consistency tools, AI Canvas for editing, and one of the most generous free tiers available — 150 images per day. Acquired by Canva in 2024. Used by over 30 million creators.
Leonardo.ai is an AI image generation platform particularly strong for game development, concept art, product visualisation and any workflow requiring consistent visual output across many images. Where tools like Midjourney excel at single striking images, Leonardo's strengths are volume, consistency, and control — producing large numbers of images in a consistent style with precise control over settings.
The platform launched in 2022 with a focus on game asset creation. It built a model library where the community trains and shares specialised models for different styles — from photorealism to anime to concept art to 3D renders. You choose the model that fits your aesthetic, set your parameters, and generate. This model selection is what makes Leonardo different from single-model tools: you are not fighting a general-purpose model to produce a specific style, you are using a model already trained for that style.
Canva acquired Leonardo AI in July 2024, bringing additional resources and the integration of Leonardo's capabilities into Canva's design platform. The standalone Leonardo platform continues to operate independently with its own features and community.
The free tier stands out: 150 images per day is substantially more than most competitors offer. For users testing the platform or doing light creative work, this free tier is a genuine production tool.
Text-to-image with model selection — Choose from dozens of community-trained models optimised for specific styles (Phoenix for photorealism, Anime Pastel Dream for illustration, Absolute Reality for photographic output, DreamShaper for artistic work). Each produces different aesthetic results. Model selection is the primary way to control output style.
Custom model training — Upload 15–20 reference images to train a custom model that generates images in your specific visual style, character or product design. Useful for brand consistency, game character sheets, or replicating a specific artistic look.
AI Canvas — An editing tool for refining generated images: inpainting (edit specific areas), outpainting (extend the image), sketch-to-image (draw a rough sketch and have AI refine it into a polished image).
Image guidance — Use an existing image as style or composition reference. Describe what you want while pointing to a reference image for the overall aesthetic or layout. Useful for producing variations on an existing design direction.
Character consistency — Generate the same character across multiple images and scenarios while maintaining consistent appearance. Particularly valuable for game development, webcomics and brand mascots.
Game developers generating assets — characters, environments, weapons, UI elements — at volume. Concept artists producing multiple variations of characters or scenes for client review. Digital illustrators building a large portfolio or working on a sustained creative project. Marketing teams producing high-volume branded image content. Anyone who needs consistent visual style across many images rather than a single striking output.
Yes — generously so. The free tier includes 150 images per day using the free model tier. Some premium models consume tokens from a monthly allocation. The token system means very heavy use of premium models requires a paid plan, but for standard usage the free tier is substantial. Artisan is $24 per month billed annually for 8,500 monthly tokens and access to all models including premium ones.
Create a free account at leonardo.ai. The free tier is active immediately with 150 daily images. The dashboard shows your generation history, model library and training tools.
This is Leonardo's most important starting step. The Platform Models section lists Leonardo's own models. The Community Models section lists user-trained models. Browse the featured models and look at the example outputs. For photorealistic images, choose Phoenix or Absolute Reality. For game assets and concept art, try Leonardo Diffusion or DreamShaper. For anime/illustration, browse the anime-focused models. The model determines 70% of your output aesthetic.
Leonardo uses standard Stable Diffusion-style prompting. Describe what you want in detail. Add style descriptors: "hyperrealistic", "cinematic lighting", "4K", "detailed". Use negative prompts to exclude unwanted elements: type in the Negative Prompt field "blurry, low quality, watermark, text, deformed".
Leonardo gives you more parameter control than most consumer tools. Key settings: Guidance Scale (how closely the AI follows your prompt — higher = more literal), Steps (number of refinement iterations — more steps = more detail but slower), Image Size (various presets and custom), Number of Images (generate 1–8 at once). Start with defaults and adjust based on results.
For any generated image that needs adjustment: click Edit in AI Canvas. Use the inpainting brush to select areas to regenerate. Describe what should replace that area. This is faster than regenerating the whole image when only one element needs changing.
Spend time in the model library first. The single biggest improvement to your outputs is choosing the right model for your style. Browse the featured community models, look at their example outputs, and test the same prompt in 2-3 different models. You will find the model that matches your aesthetic significantly improves results over the default.
Use Image Guidance for style consistency. Upload a reference image of the visual style you want and use it as Image Guidance alongside your text prompt. This is more reliable than trying to describe a style in words, especially for complex artistic aesthetics you want to replicate across multiple generations.
Train a custom model for brand or character work. For any project requiring consistent visual output across many images — a character who appears throughout a game, a product that needs to look the same from every angle, a brand mascot that needs to be usable across all content — train a custom model on 15–20 reference images. The resulting model produces consistent output far more reliably than prompt engineering alone.
Leonardo.ai was founded in 2022 and acquired by Canva in July 2024. As of 2026, the platform serves over 30 million users according to multiple industry reports. Leonardo builds on Stable Diffusion architecture with additional fine-tuning and custom model infrastructure. The platform's community model library contains thousands of user-trained models covering a wide range of visual styles.
Leonardo uses a token system where different generation operations consume different amounts of tokens. Free users receive a daily token allocation. Paid plans receive larger monthly allocations. Premium models, higher resolution generations and certain AI Canvas operations consume more tokens than standard generations. Specific credit consumption rates vary by model and settings — check the Leonardo documentation for current rates.
Leonardo's custom model training (called "Dataset Training") lets users fine-tune a model on their own images. The training requires 15–20 reference images, takes approximately 20 minutes, and produces a model that generates new images in the learned style. Per Leonardo's official documentation, trained models can be kept private or shared with the community.