Leonardo AI for beginners: understand models, Flow State and token use before you generate

OpenAI-generated editorial collage of landscape idea cards and plain counters; not Leonardo output or an interface.

Treat Leonardo AI’s model choice, generation mode and token balance as separate decisions. Choose a model and mode that fit the job, check the cost shown for that operation, and save useful results before starting another stream of ideas. In Flow State, the official help page says tokens apply to generated images, not only to the images you keep.

This is a documentation-grounded guide reviewed on 14 November 2026. We did not use a paid Leonardo account or run a hands-on comparison for this article. Interface details and plan limits can change, so use the current help pages and the information shown in your own account before committing credits.

The header is an original OpenAI-generated editorial illustration of choosing ideas within a budget. Its landscape cards are not Leonardo outputs, a screenshot of Flow State, or evidence of any model’s visual quality.

Leonardo is a platform, not one fixed image model

The official model catalogue lists models from Leonardo and other providers. Examples in the reviewed catalogue include Lucid Origin, Lucid Realism and Phoenix, alongside several partner models. The page itself calls the list a snapshot rather than a comprehensive inventory. A model available through the API also does not automatically establish the same access in every app account or plan.

That distinction matters when following tutorials. A result produced with one model and feature combination is not a reliable recipe for a different selection. Read the model name, mode and any compatibility notes before copying settings. The phrase made in Leonardo describes the platform, but it may leave out the most important information about the actual generation.

Use the official model catalogue as a starting point, then check the individual model documentation and the live selector. This article does not nominate a universal best model or score quality from vendor sample images.

Decide whether you are exploring or making a specific asset

A loose brief such as finding a visual direction for a travel newsletter is different from producing one approved header at an exact shape. Exploration benefits from varied ideas. Production needs a clear subject, constraints and a review process. Know which stage you are in before generating more pictures.

Leonardo’s help centre describes Flow State as an ideation tool that streams images and lets you narrow toward a preferred style. It contrasts that with normal Image Generation, which offers model and guidance choices for more controlled creation. That is a useful distinction, but it does not make Flow State a guarantee of a finished asset or of consistent characters.

Write a short acceptance brief before you begin. For example: one recognizable mountain landscape, a quiet sky for separately added text, and colours that fit the newsletter. This is an editorial recommendation, not a tested Leonardo prompt. The brief gives you a reason to stop when you have something suitable.

What Flow State does

The official Flow State guide describes starting with a prompt, browsing generated ideas, and using More Like This to create a new stream based on a selected image’s style. It also documents style choices for vibe, lighting, shot type and colour theme. Those controls help exploration; they are not a substitute for reviewing the output.

Save promising images before changing the prompt or settings. The help page says changes can restart the stream, and saved images appear in the Library. Keeping a shortlist avoids trying to remember which earlier result had the useful composition while a new set fills the screen.

More Like This means another generation action. It is not simply a filter over pictures you already created. If you only need to inspect, download or share an existing saved image, distinguish those actions from asking the system to produce new ones.

Read token use before treating the stream as free browsing

At the review date, the Flow State help page states a cost of one token per generated image and says that More Like This generates a new set whose images also consume tokens. That is a dated vendor statement, not a claim that every Leonardo operation costs one token. Different modes and models need their own cost check.

The same help page describes controls intended to limit generation to small groups, but its sections disagree about the on/off direction of the Scroll to Generate toggle. Do not rely on a copied toggle instruction. Read the current interface and confirm what will trigger more images before scrolling or restarting a stream.

Viewing, sharing or downloading previously generated images is described as not consuming additional tokens. The distinction is useful: the number of files you save is not the same as the number of images you asked the system to generate. Budget for exploration as well as the eventual keeper.

Set a personal stopping rule before a session. You might decide to explore one visual direction, save a shortlist and pause for review rather than continue indefinitely. That is a process recommendation, not a promise about a particular plan allowance. Check the balance and any displayed job estimate in your own account.

Plan tokens, API access and top-ups are different checks

Do not assume a subscription allowance covers every route or feature you see in a tutorial. Confirm the terms for the app or API you intend to use, the renewal rules and the feature’s current cost. We are not publishing a subscription price table here because no current pricing page was successfully verified for this guide.

The official top-up token help page describes additional token purchases for paid users and says an active subscription is required to purchase and use them. Its details should be checked before any purchase. Running out of a session budget is a good point to review the work, not an instruction to buy more automatically.

For a team, agree who can authorize spending and what counts as a finished asset. A larger token balance does not resolve an unclear brief. Keep a record of the selected model, settings and any cost shown so the next person can understand the decision without repeating the entire experiment.

Check feature compatibility instead of collecting settings

Leonardo’s documentation explains that compatible Elements depend on a model’s base. That means a list of attractive features is not a promise that all of them work together. Verify the exact combination in the current documentation and interface. If a control is absent, changing the prompt cannot make an unsupported feature appear.

For basic prompt work, our existing Leonardo image guide covers descriptive prompts and aspect ratios. The present guide addresses a different problem: deciding how model choice, ideation and token consumption fit together before you spend time generating.

If you are investigating training rather than ordinary generation, the older Leonardo custom-model article provides historical context. Verify current training support and terms rather than assuming an older feature announcement describes today’s account.

Review the image before calling the task finished

Check the selected result at its final display size and at full detail. Look for distorted objects, unreadable generated writing and unwanted marks. A pleasing exploration thumbnail may not survive a wide crop or a large presentation slide. Keep important wording in your layout tool when it needs to be exact.

Review privacy and usage terms before uploading a client reference or private photograph. Saving a file does not establish clearance for a person’s likeness, a brand design or someone else’s source image. Record permissions with the project and distinguish conceptual illustration from documentary evidence.

Prefer to generate locally?

If you have compatible Windows or Linux x64 NVIDIA hardware, Logen Free is one option for local still-image generation with a FLUX.1-schnell starter workflow. Local runs do not use per-image cloud credits, but they need suitable hardware, storage and electricity. Setup needs downloads, and application or model terms still apply.

This is not a claim of feature parity with Leonardo or of equivalent output. Cloud tools can suit people who want hosted models or do not have a compatible GPU. Compare the actual workflow you need rather than assuming one product replaces every model, editing feature and ideation mode on another platform.

Quick answers about Leonardo models and tokens

Does Flow State charge only when I save an image?

No. The official Flow State help page reviewed on 14 November 2026 says token use applies per generated image, including new images created through More Like This. It describes viewing or downloading previously generated images as not consuming additional tokens. Check the current interface because costs and controls may change.

Is every model on Leonardo the same generator?

No. Leonardo’s official catalogue lists multiple Leonardo and partner models, and identifies that list as a snapshot rather than a complete inventory. Check the exact model, generation mode, feature compatibility and account access before following a tutorial. A platform name alone does not establish which model produced a particular image.

Get the free local AI image generation guide to compare the practical requirements of a local workflow with hosted image tools.

Free local AI guide

FREE 11-PAGE PDF HANDBOOK + LOCAL AI FIELD NOTES

Get the Free 11-Page Local AI Image Generation Handbook

See what GPU VRAM you need for FLUX.1-schnell and SD 1.5, how checkpoints and LoRAs work, and five repeatable workflows on Windows and Linux x64. Enter your email to open the 11-page PDF right away and get our six-part weekly local AI field notes.

6–24 GB VRAM Tables Cloud vs. Local Math 5 Studio Workflows
How signup uses your email

We register your email with the TitanXT newsletter on beehiiv to send the 11-page PDF link, our six-part weekly local AI sequence, and occasional Logen updates. You can unsubscribe in one click from any email or by writing to support@titanxt.io. Read beehiiv's privacy policy (new tab).

Scroll to Top