Generate Makoto Shinkai style scenes with AI

Three things: near-photographic background detail, stratus-cloud skies that act like a character, and saturated blue-to-orange color grading lit by a single dominant source. On top of that sit recognizable Shinkai locations (suburban level crossings, Yotsuya-style stations, classroom rooftops) and quiet two-shot framing where weather and light carry more weight than dialogue.

Use 'stratus-cloud volumetric sky', 'anamorphic lens flare', and 'crepuscular rays through clouds' as hard requests. Pin the time of day (golden hour, blue hour, post-rain dusk) so the model commits to one direction of light instead of averaging into flat noon. If you want the train-glass reflection trick, add 'step-printed train glass reflection' and place the camera inside the carriage.

No, this is a deliberate gap. Backgrounds tend toward photoreal detail, while characters keep mainline anime proportions with layered hair and reflective eyes. That contrast is part of why the Shinkai style reads as cinematic: the figures look hand-drawn against an environment that looks observed. If you push the character art too realistic, the frame loses the Shinkai feeling and starts to look like a CG render.

Anything that lets the sky and architecture do half the work: train platforms at golden hour, rooftop two-shots before a storm, level crossings at dusk, classroom windows in summer, fireworks over a distant suburb, and meteor or rain scenes. Action posters and tight close-ups fight the style, because the Shinkai look is built around space, distance, and one quiet character moment inside a bigger frame.

Place characters small in the frame, separate them physically (across a platform, across a street, on opposite sides of a level crossing barrier), and let weather do the talking: drifting snow, summer rain, falling petals, or a meteor trail. Pair this with golden or blue hour lighting and a low horizon line. Avoid grinning poses and high-saturation backgrounds, since both undercut the quiet longing the style depends on.

Flux Dev is the default on this page because it handles dense urban detail, soft cloud volumetrics, and saturated grading without flattening the highlights. SDXL anime models can also produce a Shinkai look, but they tend to over-sharpen the foliage and lose the painterly cloud texture. Start with Flux Dev, raise the sampling steps if the sky looks noisy, and lower CFG slightly if the colors clip.

Anifusion grants you full commercial rights to images you generate, including Shinkai style outputs. The visual language (stratus skies, lens flare, level crossings) is a stylistic vocabulary and is not protected, but specific characters and titles such as Mitsuha, Taki, or Tenki no Ko are. Build your work around original characters and original locations to keep client and merchandise use safe.