If I don't, do you think it's worth a thousand bucks to get these pictures taken by the hipposter
Actually, they're all AI-generated. This is to teach you to build an exclusive photographer, Agent, to take any picture of the hippos, the real ones, and so on
Core process
We're fromOrganisationSome nice pictures begin to learn how to build photographer Agent. The photo can be summarized as follows:
- Send original map
- Send the nice picture you found Photographer Agent.
- Extract Scheme
- Agent will automatically complete the Photo Reverse Project, and export you a copy of the photo, the focus, the emotional details JSON SHOOT PROGRAM.
- One key
- PUT THIS JSON SCHEME + You want to produce a picture of the person Feeding a living mapping tool (e.g. Nano BananaPro, Bean Pack, etc.) and leaving the rest to AI just wait for the miracle。
Teach you how to open up pictures of people in photographers' ways, build photographer Agent, build your own exclusive photographers
Starting with the reset
Before we begin to formally build the cameraman Agent, let's start with the "Rock photos." It'll make you understand why you need Agent。
Assuming we see a nice picture (the following is from Little Red Bookmaker @Misty). Delete:
THERE ARE THREE WAYS TO RECREATE THIS PICTURE WITH AI。
Mode 1: original + personal photo + face changePrompt word
Tips:
Use image 2 to create the same style as picture 2 using image 1 character
advantages and disadvantages: YesMaximum Rewrite Style;but personal face referenceVery unstableNot even the person's facial information is always consulted。
Once it's drawn, it's good. (To make it easier, a photo of Liu Yofi is used here. Aggression
AND HERE'S THE EFFECT OF USING AN AI VIRTUAL MODEL. TO BE HONEST, I CAN'T TELL IF PEOPLE'S FACES ARE CROSS-REFERENCED BECAUSE THEY'RE NOT VERY OBVIOUS, BUT BOTH THE ENVIRONMENT AND THE FISSURE ARE QUITE RIGHT。
MODE 2: ORIGINAL + REVERSE PHRASING - > JSON, THEN JSON + PERSON PHOTOS
Reversed:
USE JSON TO DESCRIBE THE STYLE OF THIS PICTURE, THE MORE DETAILED IT IS。
Raw picture cue words:
{
"Systems": "Photos of women in the Queen of the Sun",
Light: {)
"Types": "natural soft and side-reverse light",
"Specifics": "Emphasis of hair and contours to create warm and luminous light.",
“Time”: “afternoon or dusk”
}
…
}
advantages and disadvantages: Face is more stable than option 1, but onlyIt's rewritten. It's not rewrittenAnd easily hallucinating。
IT'S BECAUSE WHEN THE A.I. SOLVES THE PICTURE, IT HAS ITSELFRandomWITH A RANDOM JSON, OF COURSERandom.
BUT IT'S GOOD THAT JSON HAS A STRONG RESTRAINT, AND THE PERSON WHO COMES OUT IS MORE STABLE。
Mode 3: Original + Photographer Agent-> JSON, used for the JSON + character photos
Raw picture cue words:
{
"generation_intervention": {
"intent": {
"movision": "othereal Nature Portrait / Mori-kei Steele.",
“Aesthetic_era”: “Contemporary Papanese Naturalistic Photophy,
"sention," "Gentle, dreamy, innocent, soft, fleering.",
…
}
advantages and disadvantagesThe person's face, photo details are very stable. And this Agent is all-powerfulOnce and for allIt's a real engineering program。
There are advantages and disadvantages to each of these three approaches and we can be flexible in our choice. It can even be used in combination
Think like a photographer
Method 3 was drawn up during the recapturing of the small fort. The core secret of people's pictures is photography:
Lighting:Taste and breathing
We should not just write "bright" but, more specifically, Rembrandt Lighting, Rim Light。
PETTYR: THE DIRECTION OF LIGHT DETERMINES THE 3D SENSE OF THE FACE, AND THE SIDELIGHT REMOVES THE "AI PLASTIC SENSE" MORE THAN THE FLAT LIGHT。
Camera: It's the imagePerspective and spatial sense
we should be clear on the focus: 85mm (classical human image) brings with it a vagaries of context; 35mm (human writing) contains more environmental stories。
Moood: The man who made the pictureSoul
we should not write "beautiful woman," but we control it with micro-expressions:"Surprised look back","Loose hair blowing in the wind".
This is how a professional photographer thinks
This professional way of thinking can be summed up in AI, and we'll make this into a photographer Agent。
Production photographer Agent
Don't you know anything about photography? It's okay, let AI write for you, throw this text to AI, and you can create a photographer, Agent。
I'm working on an AI Agent, a reverse project, and I'm going to enter a photo of a person (possibly a puzzle, possibly a multi-person presence), and I'm going to ask Agent to output a text that will then be used to create another person (or persons) to produce a similar picture, and Agent is going to ask:
1. PHRASING BY JSON
2. Photo details describing each cell (in the case of a multiuterine puzzle)
3. Details of the person ' s position, expression, movement, eyes, etc
4. Description of camera prop, background etc
5. Don ' t appear to be the person ' s sex, age, weight, clothing, etc
6. Write camera parameters (if they are real photographs)
7. If there is a text in the original map, identify whether it is a watermark or an art word, save only an art word and not a watermark; if there is an art word, describe as much as possible its characteristics, fonts, colours, so that it can be repeated
8. Describe the size of the original map and avoid regraving images that are deformed
9. AI Agent ' s hints are best achieved by complying with the metaphor engineering standard
10. AI Agent needs to include a process of reflection to achieve optimal results
11. There is a need to indicate the number of persons to be entered (because there may be multiple tasks in the original map) and to describe how to distinguish the person and how to facilitate the uploading of the image in a user-to-peer (which can include a description of the person at this time and avoid the error of the image by the user) by serial number of each individual and avoid the error of the picture。
If there's anything else you can add, it's to make this text, plus another image of the person, recapitulate the original picture
Please give me this AI Agent hint。
Output
# Role
You are an expert **Image Reverse Engineering Specialist** and **Prompt Architect**.
…
IT CAPTURES ALL THE PHOTOGRAPHIC INFORMATION IN THE ORIGINAL MAP AND CONVERTS THE ORIGINAL IMAGE INTO A JSON TEXT。
Generate shoot guide
Steps:
Selection Tool
The use of Nano BananaPro is recommended and their understanding of the Structured Directive (JSON) is extremely high. Of course, bean buns/dreams, Z-Image is very nice。
Enter
Enter the Agent hint into the AI tool and upload the photo that needs to be copied。
GET JSON
AI WILL RETURN TO A JSON, THIS JSON IS THE SHOOT GUIDE
Generate photos in accordance with instructions
WITH JSON GUIDANCE, WHAT'S NEXT
Selection Tool
Idem
Upload Image
Upload a photo of the person you want to generate in the dialogue box。
PASTE JSON AND SEND
SEND THE SHOOT GUIDE DIRECTLY TO THE LIVING CHART AI。
Change Agent like a code: For example, if you find that the generated Agent is not well-equipped for light capture, you can ask him to modify the Agent and adjust the output JSON. That's JSON's top control charm! Programmable, iterative。
Example: New Year's Truth
Tips:
{
"generation_intervention": {
"intent": {
"movision": "Celebratory Portrait Series for Chinese New Year, Capturing Joy and Tradition.",
"Aesthetic_era": "Contemporary Technical Chance,",
“sention”: “Joyful, Free, Warm, Auspicious”,
"A vibrant and cheerful celibration of new beginnings.",
"Global_atmosphere": "Studio setting with a warm, high low and falling confeetti."
},
…
}
Example: Winter Snow
Tips:
{
"generation_intervention": {
"intent": {
“movision”: “Fashion and lifestyle portrait photography Capturing a charming winter narrate”,
“Aesthetic_era”: “Contemporal japanese photo magazine anesthesia.”,
"sention": "Joyful, Cozy, fresh, and brightly Romantic winter vibes.",
"A cinematic snapshot of a beautiful winter day with lots of light and texture.",
"Global_atmosphere," "bright daylight, cold air, actually falling now, sun flash."
}
…
}
Example: hippos
Tips:
{
"generation_intervention": {
"intent": {
“movision”: “Personal portraits series showing diverse natural expressions,
"Aesthetic_era": "Contemporary, clean, and minimoralist portraiture.",
“Sention”: “Warm, candid, and general.”,
"soft, light, and aid.",
"global_atmosphere": "Natural indor daylight, restored and capable."
}
…
}
Summarize
IT'S NOT HARD TO USE AI TO GENERATE PICTURES, BUT IT'S HARD TO BE STABLE AND GOOD。
Submitted herePhotographer AgentTHE WAY TO HELP YOU GET INTO THE FIRST STEP OF CREATING A HUMAN IMAGE FROM AI