Before this class starts, I would like to invite you to reflect on a scene:
When you close your eyes, a beautiful picture emerges in your mind
Maybe it's a sight under a neon at night, or a starship that crosses the sky. It's clear. You can even feel the humidity of the air。
But when you open the Nano Banana Pro/the dream or Midjourney, all the vivid images collapse into a few dry words: "Girls, Rain Days, Cyberpunks, Ships."。
The pictures are naturally also dry, even full of “plastic senses”。
A lot of people get into self-doubt, "Is my language bad?"
Wrong. You don't have enough vocabulary, you don't know how to translate across dimensions。
Human thinking is sensory, mobile (image, atmosphere, emotion), while the AI model (whether dream AI or the latest Midjourney) requires rational, structured (photo, material, structure, dynamics)。
TODAY, 2026, THE TOOLS HAVE BECOME EXTREMELY UNDERSTANDINGABLE. WE DON'T HAVE TO RECITE COMPLEX SPELLS, BUT WE HAVE TO LEARN TO USE FORCE. TODAY'S LESSON, I SUM UP THREE CORE WAYS OF “REVERSE OPERATION” TO TEACH YOU HOW TO USE AI'S OWN ABILITIES TO SOLVE THE PROBLEM OF “POOR WORDS” WHEN WRITING A HINT。
METHOD ONE: SENSORY IMAGING - USING AI AS YOUR "VISUAL TRANSLATOR"
Many people are unable to write their hints because they attempt to put in place one step at a time, to write professional photographic terms or optical parameters. This not only stifles inspiration, but also makes it easier to trap it because of a lack of expertise。
CORE LOGIC: DON'T FORCE YOURSELF TO DO “TRANSLATION”, LEAVE TO YOURSELF THE RIGHT TO “DISGUISE” AND GIVE AI THE TASK OF “DISMANTLING THE IMAGE”。
1. Pain point analysis
When you want to express "a feeling of loneliness but hope," if you enter a single but hopeful image directly in the prompting box, AI often produces very rigid images (e.g. a person standing on a cliff watching sunrise). This is because “loneliness” and “hope” are abstract concepts, and AI needs specific visual elements to carry them。
Prompt_other organiser
2. Operational flows (in the case of Nano Banana Pro / Midjourney)
We do not write the hints that generate the picture, but we start with a round of “inspiration dialogue”。
STEP A: TELL AI ABOUT YOUR INTUITION
Find a large-language model (e.g. Gemini, ChatGPT, etc.) and enter the original "unprofessional" ideas in your head。
Your input:
"I want a picture. It's lonely, but a little warm. It's like a man working in a big city, walking on the road at night, and it's exhausting to see the heat on the side of the road, and it feels so good. It's got to be a little filmy."
STEP B: LET AI HELP YOU "DISMANT" THE VISUAL ELEMENTS
GIVE AI A DIRECTIVE THAT IT TRANSLATES YOUR SENSES INTO “VISIBILITY LANGUAGE”。
The command template: "Please help me decipher the image from the point of view of the photographer, based on the `silent but warm' atmosphere I described. Please provide a description keyword for the following dimensions:
Lighting: What kind of light reflects this contrast
What is the relationship between person and environment
Color Palette: How do you compare cooling
Details: What specific objects can carry the `heat gas' and `treading'?”
Step C: Get inspired and generate (tool presentations) you can use the AI tool to try the bean bag yourself。
AI MIGHT GIVE YOU A SENSE OF VISION:
Shine: Chiaroscuro (Invisible contrast): Emphasis on the high contrast between the subject and the background, as well as a sense of quality as a painting。
Image: Wide Shot / Long Shot (widen corner/vision): Shows a huge urban background, small by small by-person。
Details: Teal and Orange (Cyan): City night sky, architecture, neon lights are blue/blue (Teal/Cyan); face stand lights, heat, food orange/Amber。
In the end, the hint in Midjourney / Nano Banana Pro may become:
In the middle of nowhere, a small office walks on a journey, cold metropolis at night
Do not force yourself to think of words like warm orange steam. Let's say how you feel and let AI tell you what it is。
METHOD II: INVERSE EVOLUTION -- THE MICRO-EXPRESSION THAT ALLOWS AI TO COMPLETE THE INVISIBLE
in the production of person-type images or videos, the easiest place for you to say so is “expression” and “physical language”. in your mind, you might be "triple smile, three cool, four careless" but in the hint, you can only write smile or cool face。
CORE LOGIC: SINCE YOU LOOK AT THE REAL WORLD EVERY DAY, LET AI HELP YOU REVERSE THE WORD “REALISTIC” WITH THE WORD “INDUCTION”。
1. Pain point analysis
MICRO-EXPRESSIONS AND PHYSICAL DETAILS ARE KEY TO DETERMINING THE HYSTERIA OF THE PICTURE. MANY OF THE VIDEOS GENERATED BY AI (ESPECIALLY FROM CONCH OR GOURMETS) ARE LIKE FALSE PEOPLE BECAUSE OF THE LACK OF TRANSITIONS AND DETAILS OF FACES IN BOTH “OPEN” AND “TURN” STATES。
2. Operational flows
It's called "Turson Inversion."。
STEP A: LOOKING FOR A “SUBSTITUTE” REFERENCE CHART
Don't think. I'm going to go to the movies, I'm going to go to the photo site and I'm going to find a picture of your face that's closest to what you want. Like the look of a Liang Dynasty in "Face Year."。
STEP B: USE THE TOOL'S “REVERSE” FUNCTION
Currently, mainstream tools have a strong mapping capability。
Bean bag/Chat gpt/Gemini: Upload a reference map and enter a hint "Analysis the micro-expression of the person in this picture"。
STEP C: ANALYSIS OF ADJECTIVES EXTRACTED FROM AI
THIS IS THE MOST CRITICAL STEP. SEE HOW AI DESCRIBES THIS EXPRESSION. A.I. MAY BE ABLE TO EXTRACT FROM THIS PICTURE THE WORDS THAT YOU DON'T NORMALLY EXPECT:
1. Eyebrows and forehead: the constriction
Eyes: silent words (The Gaze)
3. Lower Face: Repressive Reclusion
AI PROVIDES A COMBINATION OF:
The film features, men, piercing and melancholy obscurity, with light and wetness in their eyes, a tight eyebrow to show internal conflict, closed lips, delicate and tight jaws, repressed emotions, royally protective film photography, dramatic dimness in this way, retro-heating, shallow, fine skin texture, emotional atmosphere
copy
STEP D: COMBINING AND APPLYING
And you can apply these "advanced words" to your role。
Apply example: A man looking at climate.
Try to write: The film features, men, piercing and melancholy eyes, light and moist eyes, eyebrow locks to show internal conflicts, close lips, tight jaws, suppressed emotions, King's Majesty's film photography, dramatic dimness in this way, reverence, shallowness, sane skin texture, emotional atmosphere
PROGRESSIST: THE CONSEQUENTIAL DESCRIPTION OF THE BODY'S REACTION IS OFTEN ACCOMPANIED BY THE BODY'S REACTION WHEN THE VIDEO IS GENERATED. YOU CAN GET AI TO HELP YOU THINK:
QUESTION AI: “IN COMBINATION WITH THE ABOVE-MENTIONED EXPRESSIONS, THEY ARE OFTEN ACCOMPANIED BY A PHYSICAL REACTION, AND WHAT KIND OF PERSONAL REACTION”
THE FILM CLOSE-UP, A MAN WITH DEPRESSED EYES, BLUE EYES AND PENETRATING. HIS THROAT WAS ROLLED UP AND DOWN (HARDLY SWALLOWED) WITH A SUDDEN AND UNSTABLE BREATH, AND HIS CHEST WAS VISIBLE. HIS JAWS WERE TIGHT, HIS MOUTH WAS NUANCED AND HE TRIED TO CONTROL HIS EMOTIONS. THE LIGHT CONTRASTS STRONGLY, WANG JIA WAI STYLE。
Write a hint: By adding these biological details to the video-generated Prompt, the real sense of the image will immediately rise to a level。
Method III: Environmental narratives - “living” the world
Many students produce videos, people move beautifully, but the background is like a poster, which is what is called a "filmhouse sense" or "show." You don't think it's because the characters are not good, but because the environment is dead。
Core logic: The environment is the extension of a person ' s mood. When you don't know how to write dynamics, write about changes in the environment。
1. Pain point analysis
in the age of video generation, the dynamic word is difficult. a lot of people only write camera pan left, making the video boring。
2. Operational flows
We're using a chain reaction。
STEP A: SET A MASTER LANGUAGE ACTION
Let's say your lead is "A running girl." A girl running on the street.
STEP B: QUESTION AI “ENVIRONMENTAL FEEDBACK”
ASK YOUR ASSISTANT AI (OR SELF-QUESTIONING): "IF A GIRL RUNS ON A STORMY STREET, WHAT WILL THE ENVIRONMENT DO TO HER? WHAT HAPPENS TO THE SURROUNDING OBJECTS?"
AI LISTS POSSIBLE CHAIN REACTIONS:
The atmosphere and body: heavy and heavy rain curtains, heavy rain pressure, wind-to-back chords, wet clothes so close to the body that the contours are drawn, the horns of clothes drawn by the wind, and the rain flows through the cheeks。
Ground: Splashing water at the bottom of the feet, intense water mist outbreaks, slurry from foot-step impacting water, mud splashing to the calves, wet bitumen, blurry motion。
Light reflection of broken and stretching neon lights on the ground, LSD from rain drops, blurry urban lights, water beads on camera, film atmosphere of water plume, wet surfaces of high contrast, conflict between cold blue gray and warm car lights。
Dynamics of surrounding objects: giant fan mists of water rolled up by motor vehicles, side trees bent by wind, leaves flipped in the wind and rain spilled on the edge of the roof to form small waterfalls, water drains, a chaotic and moving street background。
STEP C: CONSTRUCT THE "ECO-"TIP
By adding these environmental interactions to the hints, your video immediately has physical laws and life。
Optimized hints:
movie senses and shots, night, heavy rain. a girl ran on a wet city street. the slow-motion low-angle lens focused on her feet hitting the deep-water pit and provoking water flowers at every step. her wet clothes were so close to her body that her hair and windsuits were soared in the wind that she was moving backwards. the rain pours down like a thick curtain. in the background, the vehicular car rolls up huge fan fog. the humid asphalt surface reflects distorted orange neon lights. high contrast, emotional atmosphere, 8k resolution, photo-level real texture。
WHY DOES IT WORK WITH WORDS? BECAUSE YOU DON'T HAVE TO FAKE A COMPLEX MOVE, YOU JUST FOLLOW THE PHYSICAL PATTERN AND DESCRIBE "A MOVES, SO B MOVES." THIS LOGIC IS LINEAR AND VERY EASILY DEDUCED THROUGH AI。
SUMMARY: MASTERING THE "NEW LANGUAGE" OF THE AI ERA
Looking back at what we're doing today, we're not teaching you to recite words, we're building a new thinking of “human collaboration”:
WHEN YOU CAN'T DESCRIBE THE ATMOSPHERE: TELL AI HOW YOU FEEL, LET AI TRANSLATE YOU INTO VISUAL PARAMETERS。
WHEN YOU CAN'T DESCRIBE THE DETAILS: USE A REFERENCE DIAGRAM TO REVERSE IT, LET AI TEACH YOU THE MISSING MICRO-EXPRESSIONS。
CAN'T DESCRIBE DYNAMICS: LET AI PLAY A CHAIN REACTION TO THE ENVIRONMENT, AND USE THE ENVIRONMENT TO MATCH PEOPLE'S MOVEMENTS。
Phrasing is never larger than spelling whose vocabulary, but it is more “directive” than spelling who。
WHEN YOU LEARN TO LEAVE TO YOURSELF AN UNDERSTANDING OF THE WORLD, AND TO OUTSOURCE THE BORING TRANSLATION TO AI, YOU WILL FIND THAT THE INPUT BOX THAT ONCE MADE YOU SCRATCH YOUR CHEEKS WILL BECOME THE DOOR TO YOUR INFINITE IMAGINATION。
In the next class, we'll use the three tricks we've learned today to do a full siebunk specter script. Let's go back and try one way and get the dream stuff from last night。