GPT Image 2 · Genshin Impact

Genshin Impact Art Generator — GPT Image 2

OC makers hand you a generic anime face. Studio keeps the real cast on-model and drops them into Teyvat — one prompt, one cel-shaded key visual, drawn by GPT Image 2.

Cel-shaded Genshin key visual: Raiden Shogun on the stone steps of Tenshukaku at night, lanterns and the city below, rain

Cel-shaded Genshin key visual: Zhongli with a teacup on a Liyue teahouse balcony at dusk, the harbour and its lanterns below

Cel-shaded Genshin key visual: Nahida in a Sumeru library with floating books, the giant tree glowing through the arched window

Cel-shaded Genshin key visual: Venti sitting on the hand of the Statue of the Seven above Mondstadt at sunset, lyre on his knee

Cel-shaded Genshin key visual of Raiden Shogun on Inazuma cliffs with a naginata, violet lightning splitting the sky — **Raiden Shogun · Genshin Impact** — Drawn by this page's own button inside a world that already held the Genshin cast: her cover came from the world, the quiet title band from the prompt.

Cel-shaded Genshin key visual of Zhongli on a Liyue terrace at dusk, sky lanterns over the harbour and gold Geo shards — **Zhongli · Genshin Impact** — Liyue at dusk: sky lanterns over the harbour, Geo shards ringing the terrace.

Cel-shaded Genshin key visual of Nahida cupping a seed on a mossy root, the giant tree of Sumeru behind her — **Nahida · Genshin Impact** — Small hero, enormous tree: the frame sells Sumeru's scale, not just a face.

Cel-shaded Genshin key visual of Venti with his lyre on a Mondstadt parapet at dawn, turquoise wind spiral around his feet — **Venti · Genshin Impact** — Mondstadt at dawn: the lyre, the windmills, a turquoise Anemo spiral at his feet.

Cel-shaded Genshin key visual of Tartaglia with twin hydro blades in the flooded Golden House, coins on the water — **Tartaglia · Genshin Impact** — The Golden House flooded: Hydro blades, coins under the water, a grin for the camera.

Cel-shaded Genshin key visual of Paimon floating over a lantern-lit Liyue street at night, starry cape — **Paimon · Genshin Impact** — A Liyue street at night: lantern light below, a starry cape above.

Cel-shaded game key visual of Gojo in a dark uniform at a mountain shrine at blue hour, blue infinity rings around him — **Satoru Gojo · Jujutsu Kaisen** — Same recipe, different cast — Gojo at a shrine, infinity rings bending the rain.

Cel-shaded game key visual of Hori with a clear umbrella on a rainy shopping street at dusk, vending-machine light — **Hori Kyouko · Horimiya** — A school cast holds too: one umbrella, wet asphalt, shop-sign colour.

Cel-shaded game key visual of Makima standing in an empty cinema aisle, a red chain of light behind her — **Makima · Chainsaw Man** — A cinema aisle and one red chain — menace without a monster.

What Studio adds to GPT Image 2

GPT Image 2 draws the frame. The world is what makes the next one match: the cast with their covers, the places with their landmarks, the element colours, the palette. The image lands as a work with its own page, beside the characters it came from — not a file in your downloads.

A cast that stays on-model — every character is a material with a cover, and the next prompt reuses that face instead of guessing it.
Teyvat you can point at — locations keep the landmark honest — Liyue's harbour, Inazuma's torii, Sumeru's tree.
One run, one work — the frame publishes beside its world with a URL you can put on a fic cover, a thumbnail or a poster.

Questions before you draw

What do I get — one image or a set?

One key visual per run, saved as a work in the world. For a roster, run it again with the next name: the world keeps the palette and the cast, so the second frame belongs with the first.

Do I have to write the GPT Image 2 prompt myself?

No. The button carries the prompt — the game's cel-shaded look, the pose, the landmark, the element as rim light, the quiet band at the top. You change the subject, not the recipe.

Will Raiden or Zhongli actually look like themselves?

Yes, as long as they live in the world: their cover image is passed as the reference on every run. That is the difference between this and an OC maker that only knows the general style.

Can I draw my own character in this style?

Yes. Describe them, or attach a sheet — the prompt matches face, hair, eyes and costume. A Genshin-style OC usually wants one element, one nation and one weapon named up front.

What size and format is the image?

16:10, wide, about 1586×992 at high quality. That is a poster shape: big enough for a fic cover or a video thumbnail, and it still crops cleanly to 16:9.

Can it draw other anime casts the same way?

Yes — the two Jujutsu Kaisen and two Chainsaw Man frames in the grid went through the same prompt with a different reference. The recipe is the lighting and the camera, not the IP.

What if the model drifts off-model?

Attach the sheet again and keep one reference per character. Two references of the same face in one world is how a cast starts to look like two different people.

What if I don't have a world yet?

Start from one sentence. Studio builds the cast and the places from it, then the first key visual; every later frame inherits them.

Draw the first key visual for your Genshin world.

One prompt, one cel-shaded frame, published next to the cast it belongs to.

Draw My Key Visual