1. Name the image job
Start with protagonist portrait, dream-picture scene, cover art, background, or icon.
Prompt examples
A stable prompt names the subject, relationship, setting, emotion, art direction, and composition.
Start with protagonist portrait, dream-picture scene, cover art, background, or icon.
Use subjects, setting, outfit, prop, light, and weather: things the model can show.
Dream-novel prompts need distance words like childhood friend, secret relationship, protected, or meeting eyes.
Use cinematic visual novel style, dark palette, soft pink light, and no text for Yumejo-style assets.
Write who appears, where they are, how close they are, what emotion the scene carries, and what style should guide the image.
For covers and cards, generate the art without text and add lettering separately.
Dream-novel scene image, original characters only, [setting], [relationship distance], [emotion], [composition], cinematic visual novel style, dark palette, soft pink light, no text.
Protagonist prompts focus on appearance. Scene prompts focus on distance. Cover prompts focus on symbolism. Background prompts focus on light and place. Icons focus on face and mood.
If output is weak, reduce the number of people, remove text requests, make the setting concrete, and focus on one visible moment.
Dream-novel scene, one original character caring for the protagonist in an infirmary, quiet tenderness, no text. Use original characters and avoid copyrighted character lookalikes. Keep the result safe for work.
Generate an AI imageFantasy ballroom scene, an original character offering a hand, fated encounter, moonlight, no text. Use original characters and avoid copyrighted character lookalikes. Keep the result safe for work.
Generate an AI imageClear prompts are better than long prompts. Include subject, setting, emotion, and style.
Yes. You can also add English style terms when useful.
AI image generation
Create protagonist art, dream-picture scenes, covers, and backgrounds with Yumejo’s AI image generator. Adjust your prompt, then generate the image.