Leonardo AI

What is Leonardo AI?

Leonardo AI is an AI-powered platform for image and video creation, designed for designers, marketing teams, game artists, and content creators. It offers both its own developed models as well as those from various third-party providers, and it integrates functions such as text-to-image generation, reference images, natural language editing, a real-time canvas, the ability to train custom models, image scaling, and video creation within a single workspace.

Leonardo is now part of the Canva ecosystem, but it can still be registered for and subscribed to independently. Canva Business and Enterprise users who meet the requirements and do not have an individual subscription plan can access the Essential tier; the specific rules regarding this integration are indicated on the account page.

Main functions of Leonardo AI

• Generate realistic images, illustrations, conceptual designs, and commercial photos through textual prompts;
• In auto or manual mode, choose either proprietary or third-party image models;
• Use images as a reference to control characters, objects, style, composition, and content;
• Use Omni Editor to modify existing images in natural language;
• Convert sketches into complete images in real time using Realtime Canvas;
• Use the Canvas Editor for partial redrawing, image scaling, and layer manipulation;
• Perform high-resolution zooming, background removal, and other post-processing tasks;
• Create LoRAs for characters, brands, objects, or styles using a small number of training images;
• Generate videos from text or images, and control the starting frame, ending frame, and motion;
• Use Blueprints to reuse image and video production processes;
• Manage creative assets through Collections, team sharing pools, and permission controls;
• Use the official API along with Python and TypeScript SDKs for automated generation.

Image generation and model selection

Users can choose from Leonardo’s own Phoenix and Lucid series, as well as third-party models such as Flux, GPT Image, Ideogram, and Nano Banana. These different models vary significantly in terms of prompt adherence, text rendering, realism, editing capabilities, and cost; moreover, one token does not correspond to one fixed image.

Omni and Canvas editing

The Omni Editor allows users to replace objects using natural language, change the environment, adjust the composition, or create multiple variations. The Canvas Editor is better suited for masking, partial redrawing, and expanding the canvas; nonetheless, the final file should still be checked in a professional design software with regard to text, logos, edges, and printing specifications.

Realtime Canvas

Realtime Canvas employs a low-latency model to convert brushes, color blocks, uploaded sketches, and suggestions into detailed images in real time. Users can adjust the Creativity Strength, Guidance, style, random seed, and local redrawing options; once they are satisfied with the result, they can enlarge it and save it to their personal gallery.

Custom LoRA and model training

Leonardo allows paid users to upload a set of images in order to train personalized models for recognizing people, products, objects, or visual styles. The official recommendation is to prepare 10 to 20 high-quality images that are consistent yet varied; in some cases, as few as 5 images can be used.

Users must have legitimate rights to the training data.

AI video generation

The video tools support text-to-video and image-to-video generation, as well as the setting of start and end frames, motion control, and various third-party video models. The proprietary Motion model allows videos to be processed in the Relaxed video queue once the balance in Ultimate is exhausted, whereas third-party models such as Veo, Kling, and Sora always require the use of tokens.

Fast Token and Relaxed Generation

Tokens are used for operations such as processing images and videos, scaling, background removal, and model training; the cost depends on the model used, the resolution, the number of outputs, and the amount of computation required. Once the Premium balance is exhausted, it is still possible to generate Relaxed-style images using custom models, while Ultimate allows for the creation of Relaxed-style videos with custom models as well.

Third-party models, Flow State, and external API inference cannot be placed in the Relaxed queue; a Token or Top-up is required each time. Tasks generated through the Relaxed queue have lower priority and lower concurrency, and the platform can also impose restrictions on abnormally frequent use in accordance with fair usage rules.

Leonardo AI price

The individual plans include Free, Essential, Premium, and Ultimate; the public monthly prices are 0, 12, 30, and 60 dollars respectively. When paid annually, a discount of about two months is applied, resulting in monthly rates of 10, 24, and 48 dollars respectively.

Prices may vary by region, tax, and app store.

Comparison of individual plan capabilities

Team and enterprise solutions

Token Rollover saves the unused subscription tokens from the previous period in the Rollover Bank, but there are limits set for each option. Cancelling a subscription results in the loss of the balance in the Rollover Bank.

Top-up Tokens purchased separately do not expire and can be used after Rollover and the current Fast Tokens.

Leonardo API billing

The API now operates on a pay-as-you-go basis; there is no fixed monthly fee, and new users are given a small amount of test balance. Users can top up the balance manually or set up automatic top-ups, and the balance remains valid; the actual cost of each request is deducted from this balance in real time.

Leonardo AI Tutorial

Generate and edit the first image.

1. Create an account and go to Image Generation;
2. Choose Auto or a specific model suitable for the task;
3. Specify the subject, environment, composition, lighting, and style;
4. Set the aspect ratio, number of outputs, and quality options;
5. After generation, a comparison is provided regarding compliance, details, and Token costs;
6. Send the best image to Omni or Canvas for editing;
7. Fix text, hands, logos, and edges;
8. Zoom in and export the final file.

Train custom LoRA

1. Confirm that you have the permission to use the training images and portraits;
2. Prepare clear images with a consistent theme but varying angles and backgrounds;
3. Remove samples that are blurry, repetitive, contain watermarks, or have incorrect labels;
4. Create a training dataset and select the type of person, object, or style;
5. Start the training and record the number of tokens consumed;
6. Compare the base model with LoRA using the same test prompts;
7. Adjust weights to check for identity drift and overfitting;
8. Set the mature model to private and restrict team permissions.

Generate images via API

1. Create an account in API Access and obtain or top up a balance;
2. Use the Pricing Calculator to estimate model and resolution costs;
3. Create a Bearer Token and store it only on the server side;
4. Submit a generation request and record the Generation ID;
5. Poll the task until it is completed and the output URL is read;
6. Download it to your own storage so as not to rely on temporary links for an extended period.
7. Implement backoff and retry for failures, timeouts, and rate limits;
8. Monitor costs and set alerts for automatic top-ups.

Who is Leonardo AI suitable for?

• Visual designer: Generates concepts and materials and performs natural language editing;
• Game and film production teams: create characters, scenes, props, and storyboards;
• Brand marketing team: Ensures consistency between the products and the style of LoRA;
• E-commerce operation: Creating product backdrops, promotional images, and short videos;
• Illustrator: Quickly explore by combining sketches with Realtime Canvas;
• Video creators: Use various models to create image-to-video content and draft footage;
• Developers: Build generative visual capabilities through APIs and official SDKs.

The advantages of Leonardo AI

• Image, editing, training, zooming, and video tools are all available on one platform;
• Both proprietary and third-party models are available, allowing for selection based on the specific task;
• Realtime Canvas provides immediate feedback by combining sketches with AI-generated outputs.
• The paid plan offers private generation and Token Rollover;
• Premium and Ultimate can be generated in part using Relaxed once the balance is exhausted;
• The team solution offers shared tokens, permissions, and asset management;
• The official API, as well as Python and TypeScript SDKs, are suitable for production integration.

Usage restrictions and precautions

• The cost of tokens varies depending on the model and settings, and it cannot be estimated based on a fixed number of images.
• Third-party image and video models always consume Tokens and do not support Relaxed mode;
• In Relaxed mode, the speed and concurrency are lower, and it is subject to fair use restrictions;
• It is generated for free and is set to public by default, allowing the content to be used by the platform and other users;
• Training custom models may replicate biases, individuals, and protected styles in the data;
• AI-generated images may still contain errors in text, hands, structure, and branding;
• There is a limit on the amount that can be held in the Rollover balance, and canceling the subscription will result in the loss of the tokens in the bank account.
• Before commercial use, it is necessary to check portraits, trademarks, copyrights, and local AI regulations.

Privacy, training, and commercial rights

Paid users can set the generated content to Private; such private content is accessible only to them and to authorized users. Leonardo will not use it to train models without written consent. Public content is available to the platform’s users and can be used to provide, promote, improve services, and conduct training.

Paid subscribers retain all rights to the privately generated content in accordance with the platform’s terms; free users receive a non-exclusive, royalty-free license for commercial use, though the platform retains broader rights of use.

Regardless of the approach used, users must ensure that the input is valid, and they are responsible for assessing the accuracy of the output as well as ensuring that it does not violate any rights.

API, GitHub, and open-source information

Leonardo offers official REST APIs, OpenAPI documentation, as well as official Python and TypeScript SDKs. The generation of these APIs can be set to be private, and the company states that the API data is not used for training purposes. The SDKs may still be in beta version; therefore, production projects should use stable versions and test any potentially disruptive changes.

The official GitHub organization makes the SDKs and a few project examples available publicly, but the Leonardo production platform, the weights of the custom models, the training data, as well as third-party models are not made available under an open-source license. Using the open-source SDK does not mean that it is possible to use the APIs for free or to deploy models on a private basis.

Frequently Asked Questions

Is Leonardo AI free?

There is a free option that allows for 150 Fast Tokens to be refreshed per day, but the generated content is public. If private generation, more tokens, or model training are required, the paid option should be chosen.

What is the difference between Premium and Ultimate?

Premium costs 25,000 Tokens per month and supports Relaxed-style custom images; Ultimate costs 60,000 Tokens per month, and it also offers Relaxed-style custom videos, more personalized models, as well as higher queue priorities.

Can Leonardo AI be used for commercial purposes?

Yes, but the rights vary depending on the level of exposure and the specific plan. Paid users have more extensive rights to private content, while free users receive a non-exclusive commercial license; all users still need to take into account the rights of third parties.

Does Leonardo provide an API?

Pay-as-you-go APIs and official SDKs are available; the balance remains valid, and a cost estimator can be used to determine the cost of each individual request.

Is Leonardo AI open source?

The platform and models are not open source. The official SDK can be viewed and used publicly, but it is merely a client for the commercial APIs.