Key visual generator

Dark Fantasy Key Visual Generator — GPT Image 2

Most dark fantasy art starts with a pile of adjectives. This starts with the frame: one hard light, everything else falling to black, and quiet space at the top for a title — painted on your own cast, in your own world.

Dark fantasy key visual of Venti on a flooded cathedral roof under a single shaft of moonlightDark fantasy key visual of Nobara in a burning village street with a hammer raisedDark fantasy key visual of Starlight glowing on a rooftop water tower above a black cityDark fantasy key visual of Yoshikawa alone on a school field lit by one open gym door
✦Key visual
Paint a key visual of Zhongli in the sealed vault of the Geo Archon
Dark fantasy key visual of Zhongli alone in the sealed Geo Archon vault, one light on a splitting stone
Zhongli · Genshin Impact — A vault, not a portrait: the figure stays small against it, and the last contract stone splits under the only light.
Dark fantasy key visual of Raiden Shogun on a broken bridge, lit by lightning from behind
Raiden Shogun · Genshin Impact — One lightning flash behind her rim-lights the bridge; the sea and the sky go black.
Dark fantasy key visual of tiny Nahida in a buried root library holding one glowing leaf
Nahida · Genshin Impact — The leaf in her hands is the only lamp, and it dies within a few metres of roots.
Dark fantasy key visual of Tartaglia wading through a flooded fortress courtyard by lantern light
Tartaglia · Genshin Impact — One lantern on a fallen portcullis, black water to his knees, a blade dripping upward.
Dark fantasy key visual of Gojo on a blacked-out subway platform under a single emergency lamp
Gojo · Jujutsu Kaisen — A dead platform, one emergency lamp, and train doors open on nothing.
Dark fantasy key visual of Sukuna in a river of black water between broken torii, lit by a brazier
Ryomen Sukuna · Jujutsu Kaisen — Brazier red across his ribs, broken torii walking back into the dark.
Dark fantasy key visual of Homelander hovering above an empty stadium at night, lit by floodlights
Homelander · The Boys — Small in an empty stadium, one bank of floodlights, the top third left for a title.
Dark fantasy key visual of Hori in a night classroom, moonlight cutting across the desks
Hori Kyouko · Horimiya — Moonlight does the work here, not a glow: one window, one long trapezoid on the floor.
Dark fantasy key visual of Miyamura sitting backwards on a chair under a green rooftop lamp
Miyamura Izumi · Horimiya — A green service lamp on a rooftop, and the city behind him left at zero.

The frame is the product.

Neta Studio keeps the cast, the places and their covers in one world, so a key visual gets painted from the world instead of from a prompt list. GPT Image 2 draws the frame. The world decides who stands in it and what they look like.

  • Your cast stays on-model — Attach a character sheet, or let the prompt use a character's cover. Face, hair and costume come from your world, not from the model's imagination.
  • Lighting you can repeat — One source, one direction, everything else black. Ten frames from one world read as one campaign instead of ten lucky accidents.
  • Poster-ready and text-free — A wide frame with the top third kept quiet. Set your title in your own font and language, or crop the frame for a store capsule.

Questions before you paint.

What counts as a key visual here?

The one wide frame that sells a story before anyone reads it: a figure inside a place, one light, room for a title. Not a portrait, not a turnaround, and not a poster with the words already burned in.

How much art direction do I have to write?

One line: who, where, and what is happening. The prompt behind the button carries the rest — poster framing, the single light source, the tenebrist finish, the no-text rule.

Will it keep my character's face?

If you attach a character sheet or a reference photo, yes — that is what it is for. Without one it uses the character's cover in the world. A character described only in words will be invented.

What size is the frame?

A wide landscape around 1500 px on the long edge, saved in the world as a work. Crop it to 16:9, 3:2 or a store capsule from there.

Why is there no title on the image?

Models still misspell, and baked-in type locks the crop. The frame keeps the top third quiet so you can set the title in your own font and language.

How long does one frame take?

In our run the first frame came back in about half a minute of generation, and a refinement adds another. A busy world queue is the slow part, not the paint.

Can I use it on a Steam page or a game banner?

That is what the wide frame is for. Check the platform's rules for AI art, then crop the frame and set the type outside the image.

Does this only work with anime worlds?

No. The four sets here are Genshin Impact, Jujutsu Kaisen, The Boys and Horimiya — game, manga, live-action satire and slice-of-life school — all painted in the same register.

Your world already has a poster in it.

One line about the scene, one button, one frame — kept in the world as a work you can remix.

×