Leonardo.Ai Review 2026: Is It Still Worth It for AI Art?

Leonardo.Ai is a web and mobile platform that turns text prompts into images and short videos. It started in Sydney, Australia, as an image tool built on open models, and it grew into a full creative suite with its own models, a Canvas editor, video generation, 3D textures and an API. Canva bought Leonardo in July 2024, when the company said more than 19 million people had used it to make over a billion images. Leonardo still sells its own plans under its own name, and its Phoenix model also powers image tools inside Canva.

The question in 2026 is not whether Leonardo can make a good image. It can, and its Phoenix model is strong on photorealism and on text inside images, the two jobs where most generators still fail. The question is whether the token system, the renamed plans and the limits on the free tier still add up to good value now that rivals like Midjourney, DALL-E and Adobe Firefly have all improved. A tool can have the best free tier in the category and still be the wrong buy if its paid plans charge you for the models you actually want.

This review gives you the full picture with prices checked on November 10, 2026, from Leonardo own pricing page and from independent reviews that checked the same page in September and October. You will see every plan, what the tokens really buy, how the free tier limits you, how Phoenix, Lucid and Flux compare with Midjourney and DALL-E, what Prompt Magic and Alchemy do, who owns the images you make, and how the API and team plans work. Search data from DataForSEO for November 10, 2026 shows about 170 people a month in the US search for a Leonardo AI review, with low competition and a cost per click of $5.55, while about 320 a month search for Leonardo AI pricing, which tells you most buyers check the price before they check the quality.

The short answer

Yes, Leonardo.Ai is still worth it in 2026 if you want a free AI image generator that you can actually use every day. The free plan gives you 150 tokens a day, and that is enough for dozens of standard images, with access to Phoenix, Flux and the Canvas editor. No other leading image tool gives you that much for free, and for a hobbyist, a student or a creator testing ideas, free Leonardo is still the best place to start.

Yes, it is also worth paying for if you make images every week and you want to own them and keep them private. Essential at $12 a month gives you 8,500 tokens, private generation and full ownership of what you make. Premium at $30 a month is the best value for regular creators, because it adds 25,000 tokens and relaxed generation, so you can keep making images on selected models after your tokens run out. Ultimate at $60 a month only earns its fee if you need relaxed video as well, or 60,000 tokens a month for heavy third party model use.

No, Leonardo is not the best pick if you want one model with one look and zero decisions. Leonardo gives you dozens of models, presets, toggles and token costs, and that choice is the price of its power. Midjourney is simpler to learn and more consistent in style, and DALL-E inside ChatGPT is simpler still if you already pay for ChatGPT. Leonardo wins on control, on editing and on price per image, and it loses on simplicity, so pick it when you want control and skip it when you want a single button.

The plan names changed in 2026, and old reviews will confuse you if you do not know this. Apprentice, Artisan and Maestro are the old names. The current names on Leonardo own pricing page, checked November 10, 2026, are Essential, Premium and Ultimate, and the prices and token allowances stayed the same through the rename. Essential is $12 a month, Premium is $30 a month and Ultimate is $60 a month, with annual billing bringing the monthly rate down to about $10, $24 and $48. One independent review lists the annual Essential rate as $9.60 a month, which is the exact 20 percent discount, while another lists about $10 a month. Treat the checkout page as the final word on the annual rate, because the two sources disagree by 40 cents a month.

Start free if you are unsure, and stay free until the free limits actually stop you. Make ten images on the free plan, try Phoenix on a photo prompt, try Flux on a fast sketch, and use the Canvas editor to fix one bad hand or one wrong word in a sign. If you finish the week with images you like and nothing you need to own or hide, you have your answer. If you need private images, or you run out of tokens by Wednesday every week, that is the moment to pay, and Essential is the plan to try first.

What Leonardo.Ai actually is

Leonardo.Ai is a browser based platform for making images, short videos and 3D textures from text prompts, reference images or rough sketches. You type a prompt, pick a model, set the size and style, and Leonardo returns a set of images in seconds. You can then upscale one, edit one in the Canvas editor, turn one into a short video with Motion, or train a small custom model on your own images so a character or a product looks the same in every new image.

The model choice is the core of the product. Leonardo own models are Phoenix 1.0 and Phoenix 0.9, plus Lucid Origin and Lucid Realism, and the preset styles include Anime, Cinematic Kino, Concept Art, Graphic Design, Illustrative Albedo, Lifelike Vision, Leonardo Lightning, Portrait Perfect and Stock Photography. Third party models sit in the same picker, including Flux Dev and Flux Schnell, GPT Image, Nano Banana and Nano Banana Pro, Seedream, Veo 3 and 3.1 for video, Kling, Hailuo, Wan and MiniMax. That range is the main reason to use Leonardo instead of a single model tool, because you pick the right engine for each job instead of forcing one model to do everything.

Phoenix is the model most people should start with. It is Leonardo own flagship model, built for prompt adherence and for legible text inside images, which matters for posters, product shots, book covers and social graphics where a sign or a label has to read correctly. It has fast, quality and ultra modes, so you can trade speed for detail when the image matters. Lucid Origin and Lucid Realism push further toward photorealism, while Flux Dev is the pick when you want a fast, open model with a different look, and Flux Schnell is the faster, lighter version for quick drafts.

The tools around generation matter as much as the models. The Canvas editor lets you paint over part of an image, extend it past its frame, or turn a rough sketch into a finished image. Flow State gives you a scrolling feed of quick variations on one idea at 1 token per image. Blueprints turn common jobs into one click templates, such as Product in Scene, Instant Brand Kit, Bulk Resize, Furnish My Space and Outfit Inspo. Elements training builds a small custom model from your own images, so you can keep a face, a product or a style consistent across a set of images.

Leonardo is built for designers, marketers, game artists, product sellers and content creators first. A shop owner can make product photos, a marketer can make ad variants, a game artist can make concept art and textures, and a social creator can make thumbnails and posters. It also works for hobbyists and students who want to learn image generation without paying, because the free tier is real. It is not a photo editor for real photos, and it is not a video editor, so use it to make new images and short clips, then finish them in a real editor if the job needs it.

Leonardo.Ai pricing in 2026

Leonardo.Ai pricing in 2026 has four individual tiers on its own pricing page: Free at $0, Essential at $12 a month, Premium at $30 a month and Ultimate at $60 a month. Annual billing is advertised as up to 20 percent off, and independent reviews that checked the pricing page in September and October 2026 list the annual monthly rates as about $10, $24 and $48, which works out at about $120, $288 and $576 a year. Prices are in US dollars and exclude tax, which is added at checkout through Stripe, and prices can vary by region and by app store, with the iPhone app listing higher prices than the web.

Free costs $0 and gives you 150 fast tokens a day, which reset every 24 hours and do not carry over. That is about 4,500 tokens across a full month if you use it daily. Everything you make on Free is public, and other Leonardo users can see it, copy it and remix it. You get basic quality settings and one personal collection, while Blueprints, ultra quality, image and video guidance, the Realtime tools, Elements and Collections are all limited. Free is a real daily allowance, not a one time trial, and that is what makes it the best free tier in the category.

Essential costs $12 a month, or about $10 a month on annual billing, and it is the first plan worth paying for. It gives you 8,500 fast tokens a month, a rollover bank of up to 25,500 tokens, private generation, 10 personal model trainings a month, 2 simultaneous generations, a 5 item queue and top up tokens. It has no relaxed generation, so once the 8,500 tokens are gone you either buy top ups or wait for next month. Essential is the plan for a hobbyist who wants to own what they make and keep it private, and for a creator who makes images a few times a week.

Premium costs $30 a month, or about $24 a month on annual billing, and it is the best value for regular creators. It gives you 25,000 fast tokens a month, a rollover bank of up to 75,000 tokens, 20 model trainings a month, 3 simultaneous generations, a 10 item queue and unlimited relaxed image generation on selected models. That relaxed mode is the reason Premium beats Essential for heavy users, because you can keep generating on Phoenix, Lucid, Flux and the preset styles after your tokens run out, at a slower speed. Premium has no relaxed video, so video still costs tokens every time.

Ultimate costs $60 a month, or about $48 a month on annual billing, and it is the plan for power users and small studios. It gives you 60,000 fast tokens a month, a rollover bank of up to 180,000 tokens, 50 model trainings a month, 6 simultaneous generations, a 20 item queue and unlimited relaxed image and video generation on selected models. The relaxed video models on Ultimate include MiniMax H3 at 480p, Wan 2.6, Hailuo 2.3 and Hailuo 2.3 Fast, and Motion 2.0 and Motion 2.0 Fast. Ultimate only makes sense if you need that video safety net, or you burn through 25,000 tokens a month on third party models that never qualify for relaxed mode.

The annual maths is simple and worth doing before you pay. Twelve monthly Essential payments cost $144, while annual billing at about $10 a month costs about $120, so you save about $24 a year. Twelve monthly Premium payments cost $360, while annual billing at about $24 a month costs about $288, so you save about $72 a year. Twelve monthly Ultimate payments cost $720, while annual billing at about $48 a month costs about $576, so you save about $144 a year. That saving only helps if you will use the plan all year, so pay monthly for the first two or three months, watch how fast your tokens run out, and switch to annual once you know the plan fits your routine.

One warning on prices is worth stating plainly. Leonardo own pricing page, as crawled on November 10, 2026, shows the plan names and the token allowances in its table, but the dollar prices did not appear in the crawled text, so the dollar figures in this review come from independent reviews that checked the same page in September and October 2026 and agree on $12, $30 and $60 a month. The monthly prices have held steady across those checks, which gives us confidence in them, but check the pricing page yourself before you pay, because AI tool pricing shifts often and the annual rate in particular varies between sources by a small amount.

How tokens really work

Tokens are the currency of Leonardo, and they decide what everything really costs. Every action you take costs tokens, and the cost depends on how much computer power the action needs. A standard image costs a few tokens, Alchemy costs more, a video costs far more, and an upscale can cost anywhere from 73 to 1,211 tokens depending on the settings. Leonardo no longer publishes a simple per image price list, and the app shows you the cost of each action before you run it, so check the number on screen before you click generate.

The token maths on the free plan is generous but not endless. At 150 tokens a day, a standard image at 2 tokens gives you about 75 images a day, while a heavier image at 4 tokens gives you about 37 a day. Some reviews put the real daily count lower, at 25 to 40 standard images, because model choice, size and Alchemy change the cost. The safe way to think about it is this: free gives you enough images for daily practice and for a small project, and it runs out fast if you turn Alchemy on for every image or you generate video.

Paid plans change the token rules in two ways that matter. First, you get a monthly allowance instead of a daily one, so you can spend 8,500 tokens in one busy weekend and nothing the next week. Second, unused tokens roll into a rollover bank, up to three months of your plan allowance, so a quiet month is not wasted. Essential banks up to 25,500 tokens, Premium banks up to 75,000 and Ultimate banks up to 180,000. Top up tokens are also available to paid users, they never expire, and they are used after your rollover and fast tokens, but they only work while you have an active subscription.

Relaxed generation is the feature most buyers misunderstand, so read this part twice. On Premium, once your fast tokens hit zero, you can keep generating images on selected models without spending more tokens, but at a lower priority than the main queue, so jobs take longer when the service is busy. On Ultimate, the same safety net covers selected video models as well. Relaxed mode starts only when your tokens are completely gone, and you cannot turn it on early to save tokens. Only Leonardo own models and preset styles qualify, and third party models like GPT Image, Nano Banana, Veo, Kling, Seedance and Seedream always cost fast tokens, as does Flow State.

The practical rule is simple. Count your tokens for a normal week before you pay, and watch which models you actually use. If most of your work is on Phoenix, Lucid or Flux, Premium relaxed mode is a real safety net that makes the $30 plan feel like far more than 25,000 tokens. If most of your work is on Veo video or Nano Banana edits, relaxed mode will never help you, and you should judge the plan on its token allowance alone, because every generation will draw from that pool. Check the token cost shown in the app for your three most common jobs, multiply by how often you do them in a month, and pick the plan whose allowance covers that number with room to spare.

The free plan: what you really get

The free plan is the best reason to try Leonardo, and it is also the plan with the sharpest limits, so know both sides before you build on it. You get 150 tokens a day, access to Phoenix, Lucid and Flux, the Canvas editor, Flow State and the preset styles, and you can make images for commercial use under a licence, which we cover in the ownership section below. For a creator testing ideas, a student learning prompts, or a small business owner making a few social images a week, free is genuinely usable, not a teaser.

The first limit is privacy. Everything you make on the free plan is public, and every other Leonardo user can see it. There is no private mode on free, and no way to hide a generation. That rules out client work, product designs you have not launched yet, and any image with a face or a brand you want to keep private. If privacy matters at all, free is a test drive only, and Essential at $12 a month is the real entry point.

The second limit is ownership. On the free plan, Leonardo owns the images you create, and you get a non exclusive, royalty free licence to use them commercially. That licence is stated on Leonardo pricing page FAQ, but one careful review found no matching clause in the Terms of Service itself, so the legal picture is less clear than the marketing. If ownership matters, as it does for a logo, a book cover or client work, pay for a plan, where ownership vests in you on creation.

The third limit is speed and access. Free generations sit in a lower priority queue than paid generations, so at busy times, usually afternoon US time, your jobs wait longer. The Realtime tools, Elements training, Blueprints and ultra quality are limited on free, and you get only one personal collection. You also cannot buy top up tokens on free, and your daily tokens do not roll over, so a day you do not use is a day you lose.

The honest test takes one week. Use the free plan every day for seven days, make the images you would actually make for your work, and note three things: how often you run out of tokens, how often a public image worries you, and how often you want a feature that free limits. If you run out of tokens twice or more in the week, or you need one private image, Essential is worth the $12. If you finish the week with tokens to spare and nothing you need to hide, stay free and save the money.

Models: Phoenix, Lucid, Flux and the rest

Phoenix is Leonardo own flagship model, and it is the one to learn first. Phoenix 1.0 handles photorealism, portraits, product shots and scenes with text well, and it follows prompts more closely than most models, so the image you describe is the image you get. Phoenix 0.9 is the older version, still listed on the pricing page for relaxed generation, and it is a fair fallback when Phoenix 1.0 is busy or when you want a slightly different look. Both Phoenix versions are available on every plan, including free, which is a big part of why the free tier feels so generous.

Lucid Origin and Lucid Realism are the newer first party models, built for high detail and for a clean, photographic look. Lucid Origin suits general image work with strong detail, while Lucid Realism pushes further toward photos that look shot on a camera. Both qualify for relaxed generation on Premium and Ultimate, so they are safe picks when your tokens run out. If Phoenix is the workhorse, Lucid is the polish, and the difference shows most on skin, fabric, metal and glass, where Lucid keeps texture clean instead of smoothing it away.

Flux Dev and Flux Schnell come from Black Forest Labs, and Leonardo hosts them inside the same picker as its own models. Flux Dev is the stronger of the two for photorealism and for prompt adherence, and it pairs well with detailed prompts. Flux Schnell is faster and lighter, and it is the pick for quick drafts, for testing an idea, or for generating many variants in a short session. Both qualify for relaxed generation, so Flux is a safe model to build a workflow around on Premium.

Third party models are the reason many buyers pick Leonardo over a single model tool. For images you can use GPT Image, Nano Banana and Nano Banana Pro, Seedream and Flux, and for video you can use Veo 3 and 3.1, Kling, Seedance, Hailuo, Wan and MiniMax. Nano Banana costs about 40 tokens per image in published figures, which shows how fast a third party model can eat a token pool. Only paid plans can use the Veo models. Treat third party models as a rental: they are powerful, they cost tokens every time, and they never qualify for relaxed generation.

Model choice is also where beginners waste the most tokens, so use a simple rule. Start every job on Phoenix 1.0, because it is fast, it follows prompts well, and it handles text. Switch to Lucid Realism only when the image needs to look like a real photo, switch to Flux Dev only when Phoenix gives you the wrong look twice in a row, and switch to a third party model only when the job needs that model by name, such as a Nano Banana edit or a Veo video. Test a new model on one image before you generate a full set, and save the models that work for each kind of job as presets, so you stop paying to rediscover them.

Image quality vs Midjourney and DALL-E

Leonardo holds its own against Midjourney on photorealism, and it beats Midjourney on text, editing and price. In one blind test reported by an independent review, a panel of designers rated Leonardo ahead of Midjourney in 54 of 100 matched prompts, with Midjourney ahead in 38 and 8 judged equal, and Phoenix won 9 of 10 head to head tests on text inside images. Another review cites a blind photorealism test where Leonardo scored first with 68 percent preference over rivals. Those are single tests with small panels, not proof for every prompt, but they match what users report: Phoenix is strong on real looking people, products and scenes, and it is clearly stronger than Midjourney on signs, labels and logos.

Midjourney still wins on style and on consistency of look. Its images have a distinct, polished feel that clients recognise, and its community, its style references and its editor make it easier to keep a series looking like it came from one artist. Leonardo can match that look with the right model and preset, but it takes more tries and more prompt work to get there. Midjourney also has no free tier at all, which is the biggest gap between the two: Leonardo lets you make dozens of images a day for free, while Midjourney starts at $10 a month for Basic, with no free plan or standing trial.

DALL-E is the simplest of the three, because it lives inside ChatGPT, and it is the best pick if you already pay for ChatGPT and you want images inside a chat. DALL-E follows instructions well, it edits images in conversation, and it is easy for a beginner. It falls behind Leonardo on model choice, on fine control, on Canvas editing, on custom model training and on video, because DALL-E is one model with a chat box, while Leonardo is a studio with many models and real editing tools. If your images live in documents, slides and chats, DALL-E is enough. If your images are the product, Leonardo gives you more control.

Price per image is where Leonardo pulls ahead for most creators. Free Leonardo gives you about 150 tokens a day, which is dozens of images, while Midjourney Basic at $10 a month gives you about 3.3 hours of fast GPU time, or about 200 images, and DALL-E has no free tier outside ChatGPT free limits. At $30 a month, Leonardo Premium gives you 25,000 tokens plus relaxed image generation on selected models, while Midjourney Standard at $30 a month gives you 15 hours of fast GPU time plus unlimited relaxed images. The two $30 plans are close on paper, and the choice comes down to style: pick Midjourney for its look and its community, pick Leonardo for its free tier, its editing and its model range.

Our own PickyAI Midjourney review covers Midjourney pricing and quality in full, and our guide to the best AI image generators in 2026 puts Leonardo, Midjourney and DALL-E side by side with the other leading tools. Read both before you commit to a paid plan, because the right pick depends on your images, not on a single test score. The short version is this: Leonardo wins on free access, on text and on control, Midjourney wins on style and on community, and DALL-E wins on simplicity inside ChatGPT.

Prompt tools: Prompt Magic, negative prompts and image guidance

Prompt Magic is the feature to use when your prompt is short and the result is vague. It rewrites your plain language prompt into a longer, more detailed prompt before generation, adding descriptors the model responds to well. It helps beginners most, because it turns a five word idea into a prompt the model can work with. It hurts advanced users most, because it adds variance that breaks style consistency across a series, so turn it off once you know how the model reads your style, and keep it off for production work where every image in a set has to match.

Prompt Magic comes in versions, with Prompt Magic V3 the most advanced, and Alchemy adds a separate layer on top. Alchemy changes the image pipeline itself, improving lighting, contrast and fidelity, while Prompt Magic changes the words you send to the model. That split matters: use Prompt Magic to fix a weak prompt, use Alchemy to polish a good image, and do not turn both on at full strength for every image, because the token cost jumps and the image can look over processed. Test each toggle on its own first, so you know which one fixed the problem.

Negative prompts are the quiet tool that saves the most regenerations. A negative prompt tells Leonardo what to leave out, such as extra fingers, blurry text, watermarks, deformed hands or a cluttered background. Write the negative prompt in plain words, keep it short, and list only the faults you actually see in your results. A long negative prompt can fight with the main prompt, so add one fault at a time, test it, and keep the faults that fix real problems. Most users need only three or four negative terms for their usual kind of image.

Image guidance, also called ControlNet style guidance, lets you give Leonardo a reference image to control the shape of the output. You can guide by edges, by depth or by pose, so a character keeps the same pose while the style changes, or a product keeps the same shape while the scene changes. Paid plans can use up to 6 image guidance references, while free is limited. Guidance is the tool to reach for when a text prompt alone keeps giving you the right idea in the wrong shape, and it is far cheaper in tokens than generating fifty variants and hoping one lands.

The prompt workflow that works best is boring and repeatable. Write one clear sentence that says the subject, the setting and the style, add one line of detail on light and mood, add a short negative prompt for the faults you keep seeing, and pick one model for the job. Generate four images, pick the best one, and change only one thing in the next prompt, such as the model, the guidance or one descriptor. Changing one thing at a time is how you learn what each control does, and it is how you stop burning tokens on random retries.

Canvas Editor, Alchemy and upscaling

The Canvas editor is the feature that sets Leonardo apart from Midjourney and DALL-E, because it lets you fix one part of an image instead of starting over. You can paint over a bad hand, replace a background, extend an image past its frame with outpainting, or turn a rough sketch into a finished image. That saves tokens on near final images, because a targeted fix costs less than a full regeneration, and it saves time on client work, where one fault should not force you to redo a good image.

Realtime Canvas is the faster version of the same idea. You draw rough shapes and brush strokes, and the AI updates the image as you draw, so you can explore a composition in seconds instead of waiting for full generations. It is limited on the free plan, and it has a daily free generation limit, but it is the fastest way to test a layout, a pose or a product angle before you commit tokens to a full generation. Use Realtime Canvas for exploration, and use the full Canvas editor for the final fix.

Alchemy is the polish layer, and it is also the fastest way to burn tokens if you use it blindly. Alchemy improves lighting, contrast and fidelity, and it offers presets such as General, Cinematic, Illustrative and Photorealistic, with a resonance control for detail and a high resolution toggle that can double the output size. Some guides report Alchemy costing 8 to 16 times more tokens than a standard generation, and the Pro Upscaler can cost 73 to 1,211 tokens depending on settings, so check the cost on screen before you turn Alchemy on. Use Alchemy on the final image, not on every draft, and you will keep most of its benefit for a fraction of its cost.

The Universal Upscaler takes a finished image up to 8K without losing detail, using generative refinement to rebuild texture as it scales. That matters for print, for large screens and for product shots where a small image would look soft. Upscaling is not free, and the cost rises with the settings you pick, so upscale only the images you will actually use at large size. A thumbnail, a social post or a slide image does not need 8K, and upscaling it is a waste of tokens.

Elements training is the tool for consistency, and it is the one most buyers overlook. You upload 10 to 20 images of a character, a product or a style, and Leonardo trains a small custom model, called an Element or a LoRA, that recreates that subject in new images. Essential allows 10 personal model trainings a month, Premium allows 20 and Ultimate allows 50, with only limited access on free. If you make a series with the same character, the same product or the same brand style, Elements is the feature that makes Leonardo worth paying for, because no prompt alone can keep a face or a product consistent across fifty images.

Motion video: what it can and cannot do

Motion turns a still image into a short video clip, and it is the simplest video tool in Leonardo. You pick an image, set a motion strength and a camera move such as pan, zoom or tilt, and Leonardo animates the scene for a few seconds. It suits social posts, product shots with a slow push in, and concept art with drifting clouds or moving water. It does not suit long scenes, dialogue or complex action, because the clips are short and the motion is simple.

Leonardo own video models are Motion 1.0, Motion 2.0 and Motion 2.0 Fast, and Ultimate adds relaxed video on Motion 2.0 and Motion 2.0 Fast, plus MiniMax H3 at 480p, Wan 2.6 and Hailuo 2.3. Third party video models in the same picker include Veo 3 and 3.1, Kling, Seedance, Hailuo and Wan, and only paid plans can use Veo. Video costs far more tokens than images, because each second needs many frames, and third party video models always cost fast tokens, so budget video separately from images and test one short clip before you plan a series.

Start and end frames are the video control that matters most. You give Leonardo a first frame and a last frame, and it builds the motion between them, which keeps a product shot or a scene change on track instead of drifting. That control is what makes Leonardo video usable for ads and product work, where a random drift would ruin the shot. Test the start and end frames on a 3 second clip first, check that the product stays in frame, and only then make the longer version.

The honest verdict on video is that it is a useful extra, not a reason to pick Leonardo on its own. If you need video as your main output, a dedicated video tool will give you longer clips, better motion and more control. If you need a short animated version of an image you already made, Leonardo Motion is fast, it lives in the same app as your images, and on Ultimate the relaxed video safety net means you can keep making short clips after your tokens run out. Judge the video tools by that test, and do not buy Ultimate for video alone unless you make short clips every week.

Commercial licensing and privacy

Ownership is the section to read before you use Leonardo for client work, because the free and paid rules are opposites. On a paid plan, you retain full ownership and copyright of the images you generate, and you can use them for commercial projects. On the free plan, Leonardo owns the images you create, and you get a non exclusive, royalty free licence to use them commercially, as stated on the pricing page FAQ. One careful review found no matching licence clause in the Terms of Service itself, so if ownership matters for a logo, a book cover or client work, pay for a plan and keep the receipt.

Privacy follows the same split, and it is easy to get wrong. Everything on the free plan is public, and other users can see it, copy it and remix it. On a paid plan, private generation is available, but you still have to switch the privacy toggle on, or your images stay public. Public images can also be used to train AI models under the Terms, while private images will not be used to train models or build new products without your express written consent. Check the toggle on every generation that matters, because the default is not always private.

The licence also changes how you should use Leonardo for client work. A paid plan gives you ownership, which is what a client expects when they pay for images, and it gives you privacy, which is what a client expects when the work is not public yet. A free plan gives you neither, so do not use free images in a client delivery, a product launch or a brand asset. Use free for practice, for tests and for your own social posts, and switch to paid before the work leaves your desk.

Refunds are the last part of the commercial picture, and the news is not generous. Leonardo does not offer a money back window, and its help centre says refunds are assessed manually by a human agent. The Terms promise a pro rata refund only in three cases: Leonardo changes the Terms, shuts down the service, or breaches its own obligations. Assume your first month is non refundable, pay monthly for the first month, and cancel at the end of the billing period if the plan does not fit, because a cancellation takes effect at the end of the current period and unused tokens expire when the plan ends.

API and team plans

The API is a separate product from the subscription, and that split catches many developers out. API usage has its own pay as you go balance, subscription tokens do not count toward API calls, and API tokens do not count toward your subscription. There is no upfront commitment or monthly fee for the self serve API, your balance does not expire, and Leonardo offers $5 of credit to start. Pay as you go includes up to 10 concurrent generations and background removal APIs, while API Custom, arranged through sales, adds custom concurrency and per model discounts.

API pricing is consumption based, and published figures put the current rate at about $0.008 to $0.02 per token depending on the model, with some guides listing enterprise API plans from about $0.002 per standard generation. Those figures vary by model and by date, so treat them as a guide, not a quote, and check the API page for the model you plan to use. The API suits apps, websites and workflows that need images on demand, such as a product tool that generates variants, a game that needs textures, or a marketing tool that makes ad images in bulk.

Team plans suit small studios that share models and assets. Team Starter costs $72 a month, which is $24 per seat with a minimum of 3 seats, and it gives the team 75,000 shared tokens a month and a bank of up to 225,000. Team Growth costs $144 a month, which is $48 per seat with a minimum of 3 seats, and it gives the team 180,000 shared tokens a month and a bank of up to 540,000, and Growth is the only team plan that can train models on team tokens. Both team plans include unlimited relaxed image and video generation, private team generations, 6 simultaneous generations and a 20 item queue, and team plans cannot buy top up tokens.

A three seat Starter team gets 25,000 tokens per seat, the same as Premium, but each seat costs $24 instead of $30, and the team also gets relaxed video, which Premium lacks. That makes Starter a better deal than three separate Premium plans for a small team that works together, as long as the team is happy sharing one token pool. A solo creator should not buy a team plan, because the 3 seat minimum means you pay $72 a month for seats you do not use. Use the API for apps and automation, use a team plan for three or more people who share work, and use an individual plan for everything else.

Who should pay, and who should stay free

Pay for Essential at $12 a month if you make images a few times a week and you want to own them and keep them private. At about $10 a month on annual billing, you get 8,500 tokens, a rollover bank, 10 model trainings and private generation, which is enough for a hobbyist, a student with a portfolio, or a small business owner making social images. Essential is the plan to try first, because it fixes the two real limits of free, ownership and privacy, for the price of two coffees a month.

Pay for Premium at $30 a month if you make images every week and you burn through tokens on Phoenix, Lucid or Flux. At about $24 a month on annual billing, you get 25,000 tokens, a rollover bank of up to 75,000, 20 model trainings and relaxed image generation, so you can keep working after your tokens run out. That covers a freelancer, a marketer making ad variants, a game artist making concept art, or a creator making thumbnails and posters every day. Track your first two months, and if your token balance never drops below half by month end, drop back to Essential and save the money.

Pay for Ultimate at $60 a month only if you need relaxed video, 6 parallel generations, or 60,000 tokens a month for heavy third party model use. At about $48 a month on annual billing, it is the plan for a power user, a small studio, or a creator who makes short video clips every week on Motion, Wan or Hailuo. If you never use video, or your work is mostly on Phoenix and Flux, Ultimate buys you tokens you will not spend, and Premium is the better pick.

Stay on free if you are learning, testing ideas, or making a few images a week for yourself. The 150 tokens a day, the Phoenix and Flux models, and the Canvas editor cost you nothing, and the only habits you need are checking the privacy toggle, which free does not have, and remembering that your images are public and owned by Leonardo. Stay free also if you are still deciding between Leonardo, Midjourney and DALL-E, because free lets you test the real tool, not a demo, before you spend anything.

Look at a team plan only if three or more people share the work. Team Starter at $72 a month for 3 seats is cheaper per seat than Premium, and it adds relaxed video, which Premium lacks. Count your seats before you commit, and check whether one Premium account with shared presets covers your workflow first, because a small team often needs fewer seats than it thinks. Our guide to the best AI image generators in 2026 and our Midjourney vs DALL-E comparison both help with the wider choice, and our Midjourney vs Stable Diffusion comparison is the one to read if you are weighing Leonardo against an open model you host yourself.

Limitations: the token system, queues and style consistency

The token system is the biggest limitation, because it makes every image cost a different amount and it hides the real price until you check the app. A standard image is cheap, Alchemy is dear, video is far dearer, and an upscale can cost more than a whole day of free tokens. That variance is fair in one sense, because heavier work needs more computer power, but it makes budgeting hard, and a creator who turns every toggle on will run out of tokens in days. The fix is to check the token cost on screen before every generation, and to save Alchemy and upscaling for the final image only.

Queue times are the second limitation, and they hit free users hardest. Free generations sit in a lower priority queue, so at busy times your jobs wait while paid jobs jump ahead. Relaxed generation is slower still, because it runs at a lower priority than the main queue, and Leonardo says concurrency and queuing may be slowed from time to time based on demand. If you need an image in five minutes for a client, free and relaxed mode are the wrong queues, and you should use fast tokens on a paid plan instead. Generate in the morning or overnight if you are on free, when the queues are quieter.

Style consistency is the third limitation, and it is the one that matters most for a series. Leonardo gives you many models, presets and toggles, and each one pulls the style in a different direction, so a set of fifty images can look like it came from five artists. Prompt Magic adds to the problem, because it rewrites your prompt each time and adds variance. The fix is Elements training for a character or a product, a saved preset for each kind of job, and Prompt Magic off for production work. Midjourney is more consistent out of the box, so if your work is a series with one look, test both tools on the same ten prompts before you commit.

The interface is the fourth limitation, and it is a fair trade for the power. Leonardo has more panels, toggles and model pickers than Midjourney or DALL-E, and a beginner can spend a week learning what each control does. That learning curve is worth it if you make images every day, because the controls are what give you the quality, but it is a real cost if you make images once a month. Use the presets and Blueprints first, learn one new control a week, and do not try to master the whole app in a day.

The Canva ownership is a limitation only in the sense that it raises a question, not a fault. Canva bought Leonardo in July 2024, and Leonardo still operates under its own brand with its own plans, while its Phoenix model also powers image tools inside Canva. That is good news for Canva users, who get Leonardo quality inside a tool they already use, and it is neutral news for Leonardo users, who keep the full app. Watch for future changes to the plans, because an acquisition can change pricing over time, and check the pricing page before each renewal.

A 7 day test plan before you pay

Day one is for setup and a first image. Make a free account, pick Phoenix 1.0, and generate four images from one prompt about a product or a scene you actually need. Note how long each generation takes, how many tokens it costs, and which of the four you would use. That first set tells you more about the tool than any review, because it runs on your prompt and your taste.

Day two is for text and for the free limits. Generate a poster with a short headline on it, and check whether the words read correctly. Then make ten more images across the day and watch your token balance fall. That pair of tests shows you Phoenix text strength and how fast 150 tokens really go on your kind of work.

Day three is for editing. Take your best image from day one, open it in the Canvas editor, paint over one fault, and extend the frame with outpainting. Note how long the fix takes against a full regeneration. If the Canvas fix saves you a regeneration, it is earning its place in your workflow.

Day four is for models and Prompt Magic. Generate the same prompt on Phoenix 1.0, Lucid Realism and Flux Dev, then run the same prompt again with Prompt Magic on. Compare the four sets side by side, and note which model gives you the look you want with the least fixing. That comparison is the model test that decides your preset.

Days five to seven are for the decision. Make one image with image guidance from a reference photo, try one short Motion clip from a still image, and check the privacy toggle and the ownership terms for the images you made. Then look at your notes, count how many days you ran out of tokens, and decide based on real limits, not on the feature list. If you ran out of tokens twice or more in the week, or you need one private image, pay for Essential monthly first, and switch to annual once you know it fits.

1. Day 1: Make a free account, generate four images on Phoenix 1.0, and note the time and token cost.
2. Day 2: Generate a poster with text on it, then make ten more images and watch your token balance.
3. Day 3: Fix one fault in the Canvas editor and extend the frame with outpainting.
4. Day 4: Run the same prompt on Phoenix, Lucid and Flux, then again with Prompt Magic on, and compare.
5. Day 5: Try image guidance from a reference photo and one short Motion clip.
6. Day 6: Check the privacy toggle and the ownership terms for the images you made.
7. Day 7: Count how many days you ran out of tokens, and decide based on real limits.