Some links on this page are affiliate links. If you buy through them, we may earn a commission, at no extra cost to you. That does not change the verdict. Read the full disclosure

Beginner16 minLeonardo AIImageUpdated 2026-10-11

Leonardo AI: your first image workflow (2026)

Leonardo.Ai bundles generation, iteration and upscale on the web. This flow takes about 16 minutes and ends with a PNG or JPG you can show in a product mock or on social — not an abandoned test grid. You start from a real format (4:5, 9:16 or 16:9), pick a model that matches the brief and describe a shot, not a vague mood.

The problem it solves

Many people write «cinematic masterpiece 8K», leave the default aspect ratio and upscale all four thumbnails. They burn tokens, download soft images in the wrong format and assume «AI» covers commercial use. The first useful flow locks the deliverable, runs a small batch, iterates on one axis and only then raises resolution on the keeper.

What you will learn

  • Scope the deliverable and aspect ratio before the first token
  • Pick a model or preset that fits the brief (photo vs illustration)
  • Write a subject, light and exclusions prompt
  • Generate a batch and mark a direction, not the noisiest thumbnail
  • Iterate with one minimal change (light or background)
  • Upscale from the feed and download from Library
  • Check your plan’s terms before client use

Before you start

  • An account at leonardo.ai with tokens available (check the current plan on the site)
  • A concrete brief: product, portrait or concept art — not «something pretty»
  • 16 minutes. If you are unsure between image engines, read alternativas-midjourney first
  1. Steps 1

    Define deliverable and aspect ratio in a note

    Before you open the generator, write one line on where the image goes (4:5 feed, 9:16 stories, 16:9 slide) and what must appear (subject, background, forbidden: invented text, hands in the foreground). That note drives aspect ratio in Leonardo; changing format later usually means regenerating.

    What you should see: A four-line note with format, subject and exclusions. Leonardo with no prompt sent yet.

  2. Steps 2

    Open Image Generation and pick a model

    Sign in at leonardo.ai and open image generation (main prompt or Image section per the current UI). In the model picker, choose one that fits the brief: the vendor documents families such as Lucid Origin or Lucid Realism and other models that rotate in news posts. For studio-lit product, favor photoreal; for editorial illustration, a more pictorial preset. Do not switch models on every regen until you have a direction.

    What you should see: Prompt bar visible, model panel open with one model selected and aligned with your note.

  3. Steps 3

    Lock aspect ratio and batch size

    In generation controls, pick the aspect ratio from your note (common presets 1:1, 2:3, 3:2, 16:9, 9:16 per vendor documentation). If you generate multiple images per request, start with 4 for a manageable first grid. Set privacy or visibility per your plan if the UI offers it before generating sensitive content.

    What you should see: Aspect ratio 4:5, 9:16 or 16:9 selected and quantity at 4 (or the minimum your tier allows).

  4. Steps 4

    Write a shot prompt, not a pile of adjectives

    In the prompt describe subject, framing, light and background in short sentences. Product example: «Glass skincare bottle on mossy rock, soft studio key light from left, dark gradient background, no label text, product photography.» Skip chains of «ultra detailed cinematic 8K». If the UI offers automatic prompt enhancement (prompt enhance), turn it off on round one so you know what you wrote; turn it on only if the vendor recommends it and you want to compare.

    What you should see: A 2–4 line prompt in clear English or Spanish, aligned with the deliverable note.

  5. Steps 5

    Generate and pick a direction, not just shine

    Hit Generate and wait in the queue. Review the batch: look for subject and light coherence, not the most saturated filter. Mentally mark one thumbnail as direction A. If all four fail for the same reason (chaotic background, warped bottle), edit one variable in the prompt — background or angle — before you raise quantity or change model.

    What you should see: A grid of up to four images in the feed. One thumbnail flagged as keeper candidate.

  6. Steps 6

    Iterate on one axis

    With direction A clear, regenerate or use variation tools the UI shows (remix, variation or a new prompt with the same aspect ratio). Change only light or only background: «same bottle, warmer key light, softer shadow». At most two extra rounds this session. If type or hands still fail, do not stack adjectives: tighten framing in the prompt or move to a still without critical copy.

    What you should see: A second batch similar to the first with one visible controlled change, or a variation on the chosen thumbnail.

  7. Steps 7

    Upscale the keeper and download from Library

    Hover the final image and click Upscale Image (or open Upscaler from the bar or side menu per guides on leonardo.ai/news). Review mode and multiplier before you confirm; Pro Upscaler offers Precise and Creative types in Help Center documentation. When it finishes, open Library if needed and use download. Name the file with format and date (`bottle-4x5-2026-10-11.png`).

    What you should see: Upscale panel with settings visible, then a file in your system downloads.

  8. Steps 8

    Note usage rights before you show it outside

    Open your plan’s terms on leonardo.ai (commercial use, attribution, content with third-party brands or recognizable people). Update the same brief note with plan, generation date and whether use is internal, mock or client delivery. If the deliverable includes someone else’s logo or a real likeness without permission, do not ship to a paid campaign even when upscale is sharp.

    What you should see: Updated note with plan, date and allowed use type from what you read on the site.

Real use cases

Product mock for a landing page

4:5, bottle or packaging without invented text. Upscaled keeper for the hero; real copy lives in HTML, not on the generated label.

Concept art before Midjourney

You explore composition in Leonardo across models; if the final look needs --sref, you pass the keeper as reference in midjourney-sref-estilo.

Still base for video

Clean 9:16, Precise upscale, download to disk. Next: kling-primer-clip or runway-primer-clip with the same framing.

Common mistakes

  • Default aspect ratio

    You generate 1:1 and crop in Figma. You lose composition and tokens.

  • Upscaling the whole grid

    Four medium images are not four deliverables. Pick a direction first.

  • Adjective-only prompts

    «Masterpiece cinematic» does not say where the light is or what the subject is. Describe the shot.

  • Assuming commercial use

    Read the current plan. An internal mock is not the same as a paid ad.

Close

Your first useful Leonardo flow locks format, model and shot prompt, runs a small batch, iterates one axis and upscales only the keeper before downloading from Library. If the brief is in-image typography, compare with Ideogram; for moodboards with style references, Midjourney remains in alternativas-midjourney.

Next steps

  1. To pick an engine with catalog criteria, open the mejor-ia-imagenes comparison.
  2. If you are moving off Midjourney, follow alternativas-midjourney with the same four-line brief.
  3. For variations and upscale with different vocabulary, midjourney-variaciones-upscale.
  4. Tool page: /en/tools/leonardo/.

Free guide

Want the 15-tools guide?

We send it free. We will also ping you when the next guide in this path goes out.

Unsubscribe whenever you want. We do not sell the list.

Tools in this guide

Read next

You may also like