Overview

Leonardo.Ai is an image and video generation platform that runs its own models (Lucid Origin, Lucid Realism, Phoenix, Motion) alongside a long list of third-party ones. On the same account you can pick GPT-Image-1, Flux Kontext, Ideogram, Nano Banana, Seedream, Kling, Veo 3 or Sora 2, which makes it a handy place to compare outputs without juggling five subscriptions.

Beyond plain text-to-image, there's a Realtime Canvas, image editing that aims to keep characters, composition and style intact, upscaling for print-size output, and features Leonardo calls Blueprints and Flow State. You can also train a personal model from a small set of images (Leonardo suggests 10 to 20) to lock in a style, product or character. It runs on the web, Android and iOS, and there's a separate developer API.

How the plans work

Everything costs tokens, and the cost of each action depends on how heavy it is. Video eats far more than images. Prices from Leonardo's pricing page in October 2026, excluding tax:

Free: 150 fast tokens a day, public generations only, basic quality settings.
Essential: $12/month billed monthly, or $10/month billed yearly. 8,500 tokens a month, a rollover bank of 25,500, private generations, 10 personal model trainings.
Premium: $30 monthly or $24 on yearly billing. 25,000 tokens, plus unlimited "relaxed" image generation on Leonardo's own models once tokens run out.
Ultimate: $60 monthly or $48 on yearly billing. 60,000 tokens, plus unlimited relaxed video on selected models (Motion 2.0, Hailuo 2.3, Wan 2.6 and others).

Team plans start at $24 per seat (Starter, shown with 3 seats) and $48 per seat (Growth). The API is pay-as-you-go. Checked October 2026 on leonardo.ai/pricing.

One detail people miss: on the free tier, Leonardo keeps broad rights to what you make, though you still get a non-exclusive licence to use it commercially. Paid subscribers own their images outright. Third-party models never qualify for relaxed mode, so they always use tokens.

Worth it for: designers, game artists and marketers who generate a lot of variations and want one tool for both stills and short clips.

Pros: a wide model menu in one place; personal model training on every paid plan; the relaxed queue means Premium and Ultimate users don't hit a hard wall.

Cons: the token system takes a while to understand; free creations are public; the heavier third-party video models burn through tokens quickly.

Alternatives

Midjourney is the obvious comparison for pure image quality. Krea also aggregates many models with real-time tools, and Freepik's AI generator pairs generation with a big stock library.