Long time no write foundation class tutorial, almost every time is directly on the dry dismantling case, basic things like the foundation of the house, the foundation is solid and stable, the house can live in peace of mind for a long time.
Whether it's text to graph, graph to video or text to generate video, it's indispensable - theAI prompt words.
The more precise the cue word given for the same thing, the more you can meet the expectations in your mind.
Some of us use "magic to defeat magic" (there is a method at the end of the article) - send the target image to the beanbag or GPT to disassemble the image to generateAI PaintingThe cue word, this works, but then I still want to go over the basics of the theory with you, and the reason for this is in my first paragraph of this article.
I also hope that my tutorials are posts that are easy to read and understand for newbies and novices.
Without further ado, let's get to the point.
I. Core elements of the prompt
(The missing "five-dimensional descriptive approach")
1. Subject clarity: what to draw? (core content)
Define the core subject of the picture, including type, form, and key features, and avoid vague representations.
❌ Error: "A girl"
✅ Correct: "16 year old girl, waist length silver hair, cat ear hairstyle, amber pupils, smiling expression, hands on cat"
Key Tip: Include specific features (age / hairstyle / movement / props) to give the subject a three-dimensional image.
2. Stylistic orientation: what painting style to use? (Art direction)
Define painting style, art genre, or reference object, AI is extremely sensitive to style words.
Classification of common style words:
Art Genre:Impressionism, Cyberpunk, Baroque, Ukiyo-e, Minimalism
Medium Type:Oil painting, watercolor, 3D rendering, line drawing, pixel art, clay animation
Reference Object:Miyazaki style, Monet light and shadow, Makoto Shinkai scenes, Manga comic book style
✅ Example: "Watercolor style illustration with soft brushstrokes and colors referencing Miyazaki's animation"
3. Scene Setting: Where is it? (Space setting)
Describe the scene, environmental elements and atmosphere in which the subject is located to enhance the narrative of the image.
❌ Error: "in the outdoors"
✅ Correct: "Early morning forest secrets, sunlight through the leaves creating patches of light, moss and wildflowers on the ground, streams misting in the distance, quiet healing atmosphere"
Key Tips
Add time of day (early morning / dusk), weather (rainy / foggy), and environmental details (vegetation / buildings).
4. Visual Techniques: How to paint? (Lighting / Composition / Color)
Controls the visual presentation of the image and determines the final texture.
Core dimensions:
sunlight and shade
"Soft diffusion, backlit silhouettes, volumetric lighting, hard light shadows, cinematic lighting"
coloration
"Morandi color palette, highly saturated clashing colors, warm and cool contrasts, vintage fawn tones"
(art) composition
"Panoramic overhead shots, half-body close-ups, diagonal compositions, centered symmetry, golden ratio"
Resolution / Texture
"8K HD, hyper-realistic detail, sharp edges, film grain."
✅ Example: "Warm backlighting with golden contour light around the edges of the subject, panoramic composition, 8K hyper-realistic details"
5. Details to add: a plus? (Exclusive characteristics)
Include unique details or constraints to prevent AI from generating "one-size-fits-all" content.
Example:
Dress Details: "White lace dress with crushed diamond embroidery on the hem and ruffled cuffs"
MATERIAL TEXTURE: "Silk fabrics are pleated and natural, metal jewelry has old scratches, and skin is translucent and shiny"
Constraints: "No background cluttering elements, avoid over-saturation, white space in the frame 30%"
II. Cue word structure template
(Universal formula applied directly)
Based on the content of our first section we got the following AI painting prompt word universal template:
[Subject] + [Style] + [Scene] + [Visual Techniques] + [Additional Details]
Now for a few examples:
[Subject] 20-year-old girl in Chinese dress, double ringlets with pearl headdress, light pink 襦襦裙, holding a fan and half covering her face, standing under a peach blossom tree and smiling; [Style] Ancient Chinese brush painting style, referencing the soft lines of Tang Dynasty ladies' drawings; [Scene] Spring garden, full of blossoming pink peach blossoms, colorful ground, rockery in the distance; [Visual Techniques] Soft light effect, warm pink main color, flower petals with dew reflections, close-up composition of the bust, 4K HD details; [Details] jacket with entwined lotus motifs, fan painted with landscape patterns, hair with a wisp of broken hair falling down the cheek.
[Subject] Blue-haired girl with double ponytails, wearing a JK uniform and holding a magic wand, standing by the classroom window; [Style] Japanese secondary anime style, referencing the character proportions of Kaguya Oniichan; [Scene] Dusk classroom, the setting sun spilling through the window onto the desks, chalk writing on the blackboard; [Visual Technique] Warm orange light and shadow, soft and cute drawing style, close-ups of half of the body, clear lines; [Details] Bow at the neckline of the uniform, the socks with lace trim, glowing tip of magic wand.
[Body] Future mecha warrior, silver armor with red energy lines, mechanical wings behind, semi-crouching fighting posture; [Style] Hard sci-fi mechanical design, reference to the mecha texture of Pacific Rim; [Scene] Doomsday city ruins, skyscraper skeletons exposed in the broken walls, purple aurora borealis in the sky; [Visual Techniques] Cold tones are dominated by the luminous effect of the energy lines, top shot perspective, 8K hyper-realistic; [ Detail] Battle scratches on armor surface, water reflection on the ground, clear structure of mechanical joints.
[Subject] White cabin by the lake, snow on the roof, smoke coming from the chimney; [Style] Watercolor illustration style, referring to Alena Tkach's healing color scheme; [Scene] Winter morning, the lake is frozen with mist, surrounded by snowy pines, and the sky is dripping with fine snow; [Visual Technique] Soft and cold tones, snowy reflections, panoramic composition, and low saturation; [Details] Warm yellow light coming from the windows of the cabin, and snow flakes with dynamic blur effect. There is snow piled up on the branches of the trees, and the snow flakes have a motion blur effect.
Third, advanced skills: make the prompt word more "understand AI"
1. Keyword order
AI usually prioritizes recognizing the first half of the cue's content, with the core elements (subject/style) at the beginning and secondary details following.
✅ Structure: core subject + core style + scenes + secondary details + technical parameters
2. Optimizing the use of "reference words"
Add specific references to artists, movies, games, etc., and the AI will mimic their stylistic features: "Stylistic references to Spider-Man: Parallel Universe's comic book subplots and color collisions"
3. Optimizing the use of "quantifiers"
Replace vague adjectives with specific numbers or ratios:
❌ Mistake: "lots of flowers"
✅ Correct: "The 50% area of the image shows cherry blossoms in full bloom with petals falling dynamically"
❌ Mistake: "Big Scene"
✅ Correct: "Panoramic composition with the subject occupying the frame 30% and the background 70%"
IV. "Magic beats magic."
Using AIextrapolate backwardsAI Painting Prompt Words Universal Template
This method works for all people whether you're a newbie or a veteran, use it when you don't know how to write a prompt!
Send the target image to the AI (Beanbag, DeepSeek, Kimi, ChatGPT) then add a command like this:
Please understand this picture and generate a similar picture descriptor that applies to the prompt words of the i.e. Dream AI text born picture. You need to describe the picture in words from all aspects such as core subject + core style + scene + secondary details (costumes, props, etc.) + technical parameters.
Then just copy the cue word that AI gives us and go directly to the raw tool to use it.
If you are not satisfied with the image, you can continue to adjust the keywords, or upload the original image for the reference image is ok~
Did you learn it? Go ahead and try it.