Hollow Knight Key Art Generator — GPT Image 2, One Prompt
One prompt, one frame: a painterly Hollow Knight-style key visual — a small silhouette, indigo fog, one amber lantern above, the dark going down below.













A cover is a promise. This one is painted.
Neta Studio is an AI world builder — one sentence gives you a world with characters, places and events, and the key visual is one thing that world produces. The same cast keeps its face across the cover, the scene and the game.
- The style holds — One palette, one ink line, one light rule on every render — Genshin, Jujutsu Kaisen, Horimiya and The Boys all come out of the same brush.
- A face you can still name — Attach a character sheet or an existing cover and the prompt asks for the same face, hair and outfit instead of a lookalike.
- A frame you can ship — 1536×960 with the subject on a third and empty space at the edges, so a storefront crop or a thumbnail keeps the silhouette.
Questions, answered.
Is this official Hollow Knight art?
No. It is a style brief, not Team Cherry's canon, and the images here are fan art made from it. For a store page or a paid game, run the same prompt on a character you own — the palette and the light rule do not belong to anyone.
What actually makes a frame read as Hollow Knight?
Three things, in this order: a small figure with a clean outline, one amber light in a cold indigo field, and a dark depth that opens below the subject. Get those three and the frame reads as this world before any detail lands.
How do I keep my own character on-model?
Attach a character sheet or an existing render and say what must not change — face, hair colour, outfit, weapon. Without a reference the model invents a lookalike; with one it keeps the silhouette you already drew.
What size should I ask for?
1536×960 is the default here, a 16:10 key visual. Steam wants 16:9 caps, social wants 3:2 — keep the subject on a third and the empty side is what you crop, not the character.
Will it paint a title or a logo on it?
Not unless you ask. The prompt forbids text and logos, so the frame arrives clean and the title goes on in layout. If you do want painted lettering, give the exact words and expect one retry for spelling.
How many renders before one is good?
Two or three. Change one thing per run — the camera height, the amount of fog, where the light sits — and keep the rest of the prompt fixed, or you will not know which change worked.
Can I use the frame for a capsule or a thumbnail?
Yes, that is what the empty edges are for. Crop from the side the subject is not on, then check the silhouette at 200 pixels wide: if the horns or the lantern disappear, the crop is too tight.
Do I need a skill or a plugin?
No. There is no key-visual skill in the directory, so the composer carries the whole brief as one prompt: subject, palette, one light, where the depth goes, no text. You can read and edit it before you send.
Paint the cover tonight.
One prompt, one frame, one work. Start from Hornet on a Greenpath bridge and change the subject.