AI 提示词库

Take the exact face from the attached reference image as the primary subject. Use the precise facial features, bone structure, eye shape, nose, lips, jawline, skin tone, and expression of the woman shown in the reference photo. Maintain her exact face with high fidelity as the core element, but redraw the entire image in the most {argument name="style" default="clumsy, scribbly, and utterly pathetic way"} possible. Create a terrible amateur {argument name="software" default="MS Paint drawing"} on a {argument name="background" default="pure white background"} using only a mouse, with jagged, uneven lines, horrible proportions, and obvious mistakes everywhere. Make the hair look like it was drawn with the pencil tool in 5-pixel strokes, messy and scribbly with random stray lines sticking out. The eyes should be slightly misaligned, one bigger than the other, with crooked eyelashes and uneven pupils. The nose should be awkwardly shaped, the mouth lopsided and poorly filled. Skin should have visible pixelation, blotchy coloring, and bad airbrush attempts. Add terrible shading that looks like it was done with the spray can tool in MS Paint, with harsh edges and color bleeding. Dress her in a poorly drawn version of the reference outfit using blocky shapes, wrong perspective, and mismatched colors. The hands and fingers must look broken and deformed, with extra or missing fingers. Pose her in a similar position to the reference but make the anatomy completely wrong — shoulders at odd angles, neck too long or too short, head floating unnaturally. Emphasize the low-quality pixel-by-pixel feel with visible compression artifacts, dithering, and 1990s Windows 95 aesthetic. The overall drawing should be vaguely similar to the reference but also not really — kind of matching but off in a confusing, awkward, and hilarious way. Proportions should be ridiculous, symmetry completely broken, colors oversaturated or completely wrong in places. Include random MS Paint artifacts like leftover selection marquees, thick black outlines that don’t match, and clumsy fill bucket overflows. The background must be completely white with no extra elements except maybe a few accidental scribbles or stray pixels. Make it look like a child or a very unskilled adult spent 10 minutes struggling in MS Paint. Prioritize maximum pathetic quality — the worse it looks, the better. Exaggerate every flaw while still keeping the core identity of the exact face from the reference image recognizable but hilariously distorted. Photorealistic accuracy is forbidden. Only crude, low-resolution, mouse-drawn MS Paint style is allowed. Maximum chaos, maximum awkwardness, maximum comedy in the terrible execution.

Using image as reference, reinterpret the scene as a small-scale realistic high quality diorama • Convert the scene into a handcrafted miniature display mounted on a solid rectangular base with realistic terrain textures. • The diorama is held in a human hand to clearly show scale. • Soft natural lighting, macro photography, shallow depth of field. • Photorealistic, ultra-detailed, clean composition. • No text, no UI, no motion blur. • ar 1:1

Create and output one picture {argument name="style" default="like Matisse's Dance"}!

A cinematic, close-up shot of a young woman lying down on a bed, resting her head on her arm with a somber, pensive expression. She has dark hair and is wearing a dusty pink ribbed turtleneck sweater. Her hand is visible in the foreground, adorned with a simple gold wedding band and a vintage gold watch paired with metallic bracelets. The lighting is soft and diffused, casting gentle shadows across her face and highlighting the texture of the grey patterned bedding. The overall mood is quiet, intimate, and melancholically beautiful.

Summarize in Thai and check for spelling errors too

Lighting: e.g., "{argument name="lighting style" default="Golden hour backlighting"}" (soft evening light) Depth of Field: e.g., "{argument name="depth of field" default="Shallow depth of field (f/1.8)"}" (professional bokeh)

{ "prompt": "Use the uploaded image strictly as the identity reference. Preserve the same person, facial structure, hairstyle, age, gender, body proportions, and likeness exactly. Do not add or remove any person. Do not change identity. Transform the subject into a single-line continuous doodle illustration.\n\nIllustration Style:\nCreate a minimalist, hand-drawn single-line doodle where the entire subject is drawn using one continuous, unbroken line. The line should feel fluid, organic, and slightly imperfect, like a pen never lifted from the paper. Style should look artistic, playful, and intentionally simple.\n\nLine Quality:\nUse smooth but slightly shaky pen or pencil lines. Thickness can subtly vary to add character. Avoid clean vector precision; lines must feel human and hand-drawn.\n\nColor & Texture:\nPrimarily monochrome line art ({argument name="line color" default="black, dark brown, or soft navy"}). Optional very light pastel watercolor blobs or crayon patches behind the subject for warmth. Coloring must remain minimal and flat.\n\nFacial Features:\nExtremely simplified facial features formed naturally from the continuous line. Dot or small oval eyes, tiny curved smile, minimal facial detail. Expression should feel calm, friendly, and charming.\n\nProportions & Form:\nStylized doodle proportions with simplified shapes. The form should be recognizable but abstracted, maintaining likeness without realism.\n\nShading & Lighting:\nNo realistic shading or lighting. Depth is suggested only through line flow and spacing. Keep everything flat and illustrative.\n\nMood & Atmosphere:\nCreative, modern, calm, and cozy. Feels like an artsy sketch from a minimalist children’s book or creative journal.\n\nBackground:\nPlain off-white or soft pastel paper background. Optional light doodle marks or watercolor stains, kept very subtle.\n\nRendering Rules:\nNo photorealism, no photo texture, no cinematic lighting, no digital painting, no anime, no manga, no 3D render, no vector art. Must look like a genuine continuous hand-drawn doodle.\n\nFinal Output:\nA minimalist single-line continuous doodle illustration with soft artistic charm, preserving the exact identity from the uploaded image in a playful, modern doodle style.", "negative_prompt": "photorealism, realistic textures, photo texture, cinematic lighting, hyper-detail, anime, manga, 3D render, vector art, oil painting, digital painting, glossy skin, sharp clean lines, realism, face change, identity change, extra people", "quality": "high", "clarity": "high", "style_strength": "strong minimalist doodle", "reference_usage": "exact face and identity preservation required" }

A crowded Beijing subway car, but all the passengers are mythological figures (such as Sun Wukong, Zhu Bajie, etc.). On the rolling display at the carriage connection, red text reads: “Attention passengers going to the Western Heaven to fetch the scriptures, the next stop is: Gao Laozhuang.” On the glass of the train doors there is a warning sign: “Forbidden to cast spells inside the carriage.”

Mozart Requiem in D minor, K.626 {argument name="piece name" default="Dies Irae"} Munch's Scream style

a single-celled organism lifeform that is inspired by a kangaroo

Create a high-quality, editorial-style photoshoot of the same man as in the reference, at a luxurious swimming pool during daytime. The scene has bright sunshine and perfectly clear weather. He is alone, posing classically and confidently for the camera as if for a fashion magazine cover. The lighting is minimal yet sophisticated, highlighting his face and outfit with soft, natural contrast. He is wearing a fashionable, well-tailored outfit suited for a summer poolside editorial, stylish, modern, and elegant. The pool water glimmers under the sunlight, with reflections adding depth and vibrance. The background is clean, with no people around, emphasizing the premium, cinematic aesthetic and polished magazine look.

Jolivet (Composer) Pachinko for Two Pianos The pianists are Jolivet and Jolivet's Doppelgänger

Aspect ratio {argument name="aspect ratio" default="3:4"}

Make a chart of the solar system for children with one interesting fact put on each planet, dwarf planet and belt.

A prompt that, when thrown to ChatGPT, which remembers you well, will output the text part.

Change the character in the first image to the pose in the second image

I created a video by specifying 4 images generated by Nano Banana Pro in the prompt so that they switch sequentially.

{ "prompt_type": "ultra_realistic_cinematic_full_body_disintegration", "base_style": "ultra-realistic, cinematic, photo-real digital painting, hyper-detailed, dramatic high-contrast lighting", "quality_tags": [ "8K resolution", "masterpiece", "best quality", "ultra high detail", "photorealistic textures", "cinematic color grading", "sharp focus", "volumetric lighting", "HDR", "realistic skin texture", "subtle film grain" ], "subject": { "gender": "female", "view": "full-body side profile facing right", "face_reference": "use provided face reference exactly with zero alterations", "facial_accuracy": "exact facial structure, skin tone, eye shape, nose, lips, bone structure, hairstyle replicated perfectly", "expression": "deep introspection and sorrow, subtle tension in eyebrows, slightly parted lips, soft emotional heaviness", "eyes": "looking downward with emotional depth", "hairstyle": "copied exactly from reference image, no changes in length, texture, or styling" }, "pose": { "description": "elegant sorrowful stance", "posture": "slightly forward-leaning posture, shoulders gently lowered", "hands": "hands positioned naturally according to pose, relaxed yet emotionally expressive", "body_language": "introspective, vulnerable, emotionally weighted" }, "outfit": { "description": "flowing, textured fabric outfit that reacts naturally to wind and light", "material_detail": "fine fabric fibers, realistic folds, subtle fabric translucency", "color": "muted neutral tones to complement monochrome environment" }, "disintegration_effect": { "coverage": "nearly half of the body dissolving", "affected_areas": [ "left arm", "left shoulder", "left torso", "left leg", "portion of hair" ], "effect_type": "thick cloud-like grey smoke", "particle_detail": "glowing yellow dust particles embedded within smoke", "transparency": "gradual transparency transitioning into luminous particle dispersion", "motion": "active disintegration mid-process, particles drifting outward dynamically", "lighting_interaction": "particles catching rim light and glowing intensely" }, "background": { "environment": "minimal desolate concrete or paved surface", "sky": "heavily overcast dramatic sky", "atmosphere": "moody, cinematic, emotionally heavy" }, "backlight_object": { "type": "massive glowing circular energy sphere", "light_color": "fiery golden-white plasma", "function": "powerful rim and backlight", "effect": "strong silhouette separation, dramatic edge lighting on smoke and particles", "plasma_detail": "radiating energy waves, subtle solar flare effects, bloom and glow" }, "lighting": { "style": "extreme high-contrast cinematic lighting", "rim_light": "intense golden-whi" } }

Man in an apartment.

With the prompt 'crop only the center' of the first image, Nano Banana Pro perfectly generated the next image.

A medium shot of the {argument name="character_count_en" default="14"} fluffy characters sitting squeezed together side by side on a worn beige fabric sofa and on the floor. They are all facing forward, watching a vintage wooden-box television set placed on a low wooden table in front of the sofa. The room is dimly lit, with warm light from a window on the left and the glow from the TV illuminating the creatures’ faces and fluffy textures. The background is a cozy, slightly cluttered living room with a braided rug, a bookshelf with old books, and rustic kitchen elements in the background. The overall atmosphere is warm, cozy, and amused.

“Please remove the overlaid text from the image”

1980s US import anime.

Listen, absolutely! Don't mix in anything! Anything different!!