Hi there đź‘‹
We are the founders of Maxfusion.ai and we spend tons of time digging up how the best models in the market perform for AI UGC ads & Marketing tasks Follow us for constant updates with 0% BS!
And if you have questions about Gemini, Nano banana or any other way to scale your ads with AI - ALWAYS REACH OUT TO US. WE ALWAYS RESPOND ASAP
| Ori Zilbershtein | Stav Zilbershtein |
| https://www.linkedin.com/in/orizilbershtein/ | https://www.linkedin.com/in/stavzilbershtein/ |
| https://x.com/OriSilver | https://x.com/StavZilber |
Unlike typical guides we won’t go over the theory of what it is and where it came from, as you came in for the secret sauce, not only you’ll get free JSON templates but an entire set of tools that generates JSON prompts for you from any image.
First we need to address the fact you suck at prompting, you don’t know this but this is the main reason you thought JSON prompt is voodoo or simply snake oil.
Why JSON Prompting is Superior
Your results are just as good as your prompt, and this is confusing to many.
You need detailed, vivid prompts for the best results, but you also need to know when to describe a "vibe" like "grungy looking" versus when specific details are needed like "worn brown leather jacket". The more details the better, and that is where people find it hard, not everyone has a vivid imagination or even the time to spend on this.
The lack of details and more importantly the lack of balance in your prompt is your main issue. That is why JSON formatting, when structured correctly, solves all these issues.
Addressing the myth
Some claim JSON prompting is voodoo, that AI image models ignore it or simply convert it to text. This misconception exists because users take their already weak textual prompts and convert them into JSON format. They're simply repackaging a bad prompt in a different structure, which produces the same poor results.
JSON Prompt Key advantages:
When you use JSON, you're essentially giving the AI a detailed blueprint rather than a vague description. This structured approach ensures each element of your image is clearly defined and properly weighted.
What you should expect?
JSON master list of elements and logic (Build your own JSON structure)
The JSON structure follows a Divide and Conquer logic, segmenting the image analysis into distinct, non-overlapping domains. The process starts with Global Context and Metadata, then moves to Aesthetic attributes (Color and Composition), and finally decomposes the scene into its constituent, granular elements: Objects and Text. This hierarchy ensures that every captured detail is logically nested under the relevant parent category (e.g., all object-specific details are guaranteed to be found within the "objects" array), creating a robust, machine-readable database record that minimizes redundancy and simplifies querying. To make the AI efficient we insert a protocol that helps eliminate the AI pitfalls when adhering to prompts and data. And it drills down in three distinct stages to extract the most accurate data possible for the JSON. Macro Sweep —> Micro Sweep —> Relationship Sweep
â–¶ This list contains all unique elements (keys) from the schema with a brief descriptor.
â–¶ This list shows the parent-child relationships using indentation.
â–¶ Core directive to give the AI on how to analyze and use the JSON structure
ROLE & OBJECTIVE
You are VisionStruct, an advanced Computer Vision & Data Serialization Engine. Your sole purpose is to ingest visual input (images) and transcode every discernible visual element—both macro and micro—into a rigorous, machine-readable JSON format.
CORE DIRECTIVE
Do not summarize. Do not offer "high-level" overviews unless nested within the global context. You must capture 100% of the visual data available in the image. If a detail exists in pixels, it must exist in your JSON output. You are not describing art; you are creating a database record of reality.
ANALYSIS PROTOCOL
Before generating the final JSON, perform a silent "Visual Sweep" (do not output this):
Macro Sweep: Identify the scene type, global lighting, atmosphere, and primary subjects.
Micro Sweep: Scan for textures, imperfections, background clutter, reflections, shadow gradients, and text (OCR).
Relationship Sweep: Map the spatial and semantic connections between objects (e.g., "holding," "obscuring," "next to").
OUTPUT FORMAT (STRICT)
You must return ONLY a single valid JSON object. Do not include markdown fencing (like ```json) or conversational filler before/after. Use the following schema structure, expanding arrays as needed to cover every detail:
CRITICAL CONSTRAINTS
Granularity: Never say "a crowd of people." Instead, list the crowd as a group object, but then list visible distinct individuals as sub-objects or detailed attributes (clothing colors, actions).
Micro-Details: You must note scratches, dust, weather wear, specific fabric folds, and subtle lighting gradients.
Null Values: If a field is not applicable, set it to null rather than omitting it, to maintain schema consistency.
the final output must be in a code box with a copy button.
A tool to adjust weights of your master prompt
With this tool you can adjust the priorities and hierarchy of the different elements, now a word of warning here. When tested with the best image models, NBP / Seedream etc. Their reasoning models are so strong that changing the different elements hierarchy did little to no effect on the final image produced Still felt this might be a useful tool to allow you to play with the prioritiies, so I included it here.
You can access the JSON architect here
Real-World Use Cases + Templates you can use right now
1. Static Ads
Create perfectly branded advertising content with consistent styling across campaigns. This is perfect for static ads for example.
If you copy the JSON and change the elements you’ll get a version of this:
{
"subject": {
"description": "A female model posing for a shapewear advertisement, body angled slightly to the right, looking directly at the camera, hand pulling left strap",
"layout_rules": "maintain advertisement layout with text on left and model on right, preserve distinct footer and headline positioning",
"age": "young adult",
"expression": "neutral, confident, and alluring",
"hair": {
"color": "blonde highlights",
"style": "shoulder-length, voluminous tight curls/coils with a center part"
},
"clothing": {
"bodysuit": {
"type": "sculpting shapewear bodysuit",
"color": "mocha brown",
"details": "spaghetti straps, scoop neckline, mid-thigh length shorts, seamless tight fit"
}
},
"face": {
"preserve_original": true,
"makeup": "natural glow, sculpted brows, nude lip color"
}
},
"accessories": {
"graphics_and_overlays": {
"sticker_badge": {
"type": "white scalloped sticker with purple outline",
"text": "BUY 1 GET 1 FREE",
"position": "center right, floating over model's torso",
"detail": "purple arrow pointing from main text to badge"
},
"color_swatches": {
"type": "5 circular color icons",
"colors": "black, dark brown, beige, white, purple",
"position": "bottom left above footer"
}
}
},
"photography": {
"aspect_ratio": "1:1",
"resolution": "1024x1024",
"camera_style": "studio e-commerce photography",
"angle": "eye-level, medium shot",
"shot_type": "commercial advertisement composition",
"texture": "high resolution, smooth skin texture, studio lighting, clean sharp edges"
},
"background": {
"setting": "studio backdrop",
"wall_color": "soft grey/white with subtle fabric folds",
"elements": [
"Top Left Text: 'too tight', 'too restricting', 'too expensive' (all in black with strikethrough lines)",
"Main Headline: 'SCULPTING SHAPEWEAR BODYSUIT' in bold purple sans-serif font",
"Sub-headline: 'Now available in 5 colors' in small purple text",
"Brand Logo: '< Maxfusion AI >' in white serif font at bottom left",
"Footer Bar: Solid purple banner at very bottom",
"Footer Elements: Brand logo, 5 gold stars, text 'RATED 4.7/5 BASED ON 335 REVIEWS'"
],
"atmosphere": "clean, professional, persuasive commercial vibe",
"lighting": "soft studio strobe lighting with gentle shadows"
}
}
2. Cinematic or unique product shots
You can create a specific scenario and simply replace the product within it by changing the product in the template. Just replace [PRODUCT] with whatever you want to create - Apple Watch, Kellogs Rice Crispies, MacBook Pro etc This will create a beautiful translucent product shots.
{
"style": "photorealistic product render",
"scene_type": "isolated object",
"object": {
"type": "[PRODUCT]",
"material": "ultra-clear iridescent glass",
"surface_texture": "glossy, smooth, refractive",
"color": "transparent with subtle rainbow refractions",
"details": "all internal components clearly visible with high structural fidelity; internal contents, assemblies, or materials are accurately represented according to the nature of the product (e.g. circuitry, packaging fill, mechanical parts). External surfaces retain authentic branding, including logos, typefaces, textures, and graphic elements in their correct placement and scale. All visual identifiers such as branding colors, mascots, and logos must exactly match the real-world product’s official design."
},
"environment": {
"background": "neutral gray or white seamless studio backdrop",
"shadows": "soft diffused shadows under and around the object",
"scale": "true-to-life proportions, matching real-world dimensions of the actual product"
},
"lighting": {
"type": "high-Kelvin studio lighting",
"direction": "multi-point, overhead and side fill",
"intensity": "bright, even exposure",
"effects": "crisp reflections, subtle caustics, high transparency definition"
},
"post_processing": {
"glow": "none",
"bloom": "minimal, only on brightest reflections",
"depth": "mild clarity boost on internal parts",
"sharpness": "high detail clarity, especially on internal mechanisms or contents"
},
"camera": {
"angle": "3/4 view",
"focus": "entire object in sharp focus",
"depth_of_field": "deep DOF, everything in focus",
"framing": "centered, product takes up 70-80% of frame"
}
}3. Contact Sheet JSON
This form of JSON is extremely powerful, the main reason to create multi shots other than cinematic aesthetics is to overcome the inconsistencies presented when trying to prompt the same style in different generations.
You might know this from experience, you get an amazing character and look & feel but when you prompt it in different scenes or other angles that “feel” or “vibe” is lost. That lost of voibe is costing you customers, we are wired to pick up on these differences and even without being able to pinpoint why something looks off, it does.
Contact sheet is the solution to overcome this by prompting multiple shots in one sheet to then seperate and generate as standalone images. (Here is our full contact sheet prompt guide)
Contact sheet is also a good way to mimic styles of famous photographers cinematographers and artists in general, putting it all in one sheet and allowing you to swap characters quickly.
Phone booth style images
Generate multiple variations with precise control over each element. This lets you create a consistent vibe while swapping out the character, for example, this photo booth style. Simply change the character using a reference image.
{
"scene":
{
"format":"collage_4_portraits",
"style":"extremely_realistic_cinematic_editorial",
"background":"minimal_plain_light_colored_wall",
"color_mode":"color",
"texture":"film_grain_vintage_polaroid",
"lighting":"direct_flash_high_contrast_harsh_shadows_slightly_overexposed_90s_2000s_editorial"
},
"subject":
{
"type":"young_woman",
"face":"do_not_change_the_face",
"beauty_level":"extremely_beautiful",
"age":"young_adult",
"hair":"long_blonde_red_with_wispy_bangs",
"eyes":"green",
"top":"black_form_fitted_halter_top_plunged_neckline",
"accessories":"layered_black_thin_choker_necklace_with_large_black_marble_heart_pendant",
"makeup":
{
"eyeliner":"thin_precise",
"lashes":"voluminous_natural",
"cheeks":"soft_contour",
"lips":"glossy_red"
}
},
"collage":
{
"arrangement":"2x2_grid",
"frames":
[
{
"position":"top_left",
"shot_type":"close_up",
"pose":"sensually_applying_lipgloss",
"gaze":"slightly_off_camera",
"expression":"confident_alluring"
},
{
"position":"top_right",
"shot_type":"close_up_mid",
"pose":"holds_hair_back_with_both_hands",
"gaze":"straight_into_camera",
"expression":"fierce_intense_stare"
},
{
"position":"bottom_left",
"shot_type":"mid_shot",
"pose":"arms_elegantly_placed_on_waist",
"gaze":"direct_to_camera",
"expression":"slightly_parted_lips_sultry_mysterious_mood"
},
{
"position":"bottom_right",
"shot_type":"close_up",
"pose":"one_hand_gently_touches_lower_lip",
"gaze":"glancing_to_the_side",
"expression":"subtle_reflective"
}
]
},
"camera":
{
"angles":"close_up_and_mid_shot",
"composition":"straight_on_centered",
"lens_style":"flash_portrait_editorial"
},
"post_processing":
{
"film_grain":"visible_medium",
"contrast":"high",
"exposure":"slightly_overexposed_for_bold_look",
"overall_vibe":"moody_vintage_90s_2000s_fashion_portrait_collage"
}
}Make up tutorial split shot
{
"format": {
"layout": "1x4_grid",
"aspect_ratio": "16:9",
"panel_order": ["Base", "Eyes", "Contour", "Lips"]
},
"subject": {
"identity_reference": "attached_woman",
"appearance": {
"hair": "dark_brown_high_ponytail",
"skin": "warm_fair_skin_with_freckles",
"eyes": "light_colored",
"accessories": ["gold_hoop_earrings"],
"outfit": "simple_black_top"
},
"expression": "neutral_soft_smile",
"camera": {
"framing": "tight_portrait",
"lighting": "soft_even_studio_light",
"background": "clean_white_background"
}
},
"panels": [
{
"title": "Base",
"description": "The woman demonstrates applying a light foundation and concealer. Smooth, natural, dewy finish. Minimal retouching. Shows a makeup sponge gently touching her cheek.",
"visual_focus": ["skin_tone_evenness", "natural_glow"]
},
{
"title": "Eyes",
"description": "She applies eyeliner and mascara with precision. Subtle brown eyeshadow blended smoothly. Shows a close-up angle where her eye makeup is clearly visible.",
"visual_focus": ["eyeliner_shape", "lashes_definition", "soft_shadow_blend"]
},
{
"title": "Contour",
"description": "She demonstrates cheek contour and highlight placement. A makeup brush is lightly brushing her cheekbone. Subtle sculpting, realistic shading.",
"visual_focus": ["cheekbone_contour", "highlight_reflection"]
},
{
"title": "Lips",
"description": "She applies a natural rosy lip color. Close-up of lips but still within portrait frame. Soft and realistic texture.",
"visual_focus": ["lip_color_application", "natural_matte_finish"]
}
],
"visual_style": {
"render": "hyper_realistic",
"color_tone": "neutral_studio",
"sharpness": "high",
"consistency": "same_model_same_lighting_all_panels",
"tutorial_labels": {
"position": "bottom_center",
"typography": "clean_modern_sans_serif",
"size": "medium"
}
}
}Araki nabouysho style photos
One of the most iconic photographers of our time, you can now create your own image in the style of his most famous images. Simply drop your image or any person together with the JSON provided to create a stunning Contact sheet image in his style.
Original:
Araki nabouysho style images:
{
"system_task": "ARAKI_CONTACT_SHEET_V2_DIRTY",
"input_processing": {
"source": "reference_image_provided",
"action": "extract_subject_identity",
"constraint": "maintain_facial_features_100_percent"
},
"global_style_override": {
"visual_imperative": "Must look like a physical, damaged, flash-lit photograph from 1970s Tokyo. NOT a digital filter.",
"lighting_physics": {
"type": "Point-Source On-Camera Flash",
"behavior": "Center-weighted illumination with heavy vignetting (dark corners).",
"shadows": "Hard, sharp cast shadows close to the subject."
},
"film_damage_layer": {
"grain_structure": "Coarse, uneven silver halide grain (ISO 1600+)",
"artifacts": "White dust specks, micro-scratches, dirty negative scan aesthetic",
"contrast_curve": "Crushed blacks (lost detail) vs. Blown-out highlights"
},
"skin_physics": {
"texture": "Oily, humid, unpowdered",
"specularity": "High specular highlights on forehead/nose/cheeks (flash bounce)",
"quality": "Raw, porous, imperfect"
}
},
"contact_sheet_composition": {
"layout": "2x3 Grid",
"frame_instructions": [
{
"frame_id": 1,
"archetype": "The Boat (Sanzu River)",
"visual_prompt": "High-angle snapshot. [Subject] curled in fetal position in a wooden boat. Water is dark and oily. Lighting: Flat, overcast melancholic light but with high contrast grain. Vibe: Rough, unstaged, lonely."
},
{
"frame_id": 2,
"archetype": "The Balcony (Daily Life)",
"visual_prompt": "Medium shot. [Subject] leaning on rusty railing. Background: Overexposed white sky (blown out). Lighting: Fill-flash hitting the face to create 'shiny' skin against the bright sky. Vibe: Domestic snapshot."
},
{
"frame_id": 3,
"archetype": "Erotos (Flash Macro)",
"visual_prompt": "Extreme Macro Close-up. Harsh direct flash. [Subject]'s face is oily and sweaty. The background is pitch black (flash fall-off). Catchlights in eyes are sharp white dots. Focus: Pores and skin texture."
},
{
"frame_id": 4,
"archetype": "Kinbaku (Tatami Room)",
"visual_prompt": "[Subject] sitting in a tatami room. Rope binding is visible but looks impromptu. Lighting: Hard flash casting sharp, complex rope shadows on the skin. Background: Cluttered, textured walls, not a clean studio."
},
{
"frame_id": 5,
"archetype": "Flower Rondeau (Decay)",
"visual_prompt": "Tight framing. [Subject]'s face next to a wilting, dead flower. The flower texture is crispy/dry. The skin is shiny/wet. High contrast. Flash lighting."
},
{
"frame_id": 6,
"archetype": "The Morning After (Messy Room)",
"visual_prompt": "Dutch angle (tilted). [Subject] lying on messy sheets. Background: Dresser with bottles, trash, clothes on floor. Lighting: Flash mixed with window light. Vibe: Intimate, dirty, voyeuristic."
}
]
}
}Cindey Sherman style Contact sheet
Copying styles and even famous scenes re-imagined in styles is extremely easy with JSON prompting because of the attention to details. Original:
Cindy Sherman Style images:
{
"system_task": "CINDY_SHERMAN_CONTACT_SHEET_V2_GRANULAR",
"input_processing": {
"source": "reference_image_provided",
"action": "cast_subject_into_roles",
"critical_constraint": "Subject must look like they are wearing prosthetics/wigs. Do not beautify. Retain facial structure but apply 'stage makeup' texture."
},
"global_style_override": {
"philosophy": "Staged Artifice. The images must look like high-resolution photographs of *fake* moments. High fidelity textures of low-fidelity props.",
"film_stocks": {
"color": "Kodak Portra 400 (Saturated, slightly warm)",
"bw": "Kodak Tri-X (High contrast, grain)"
}
},
"contact_sheet_composition": {
"layout": "2x3 Grid (6 Frames Total)",
"frame_instructions": [
{
"frame_id": 1,
"archetype": "The Grotesque Socialite (Ref: Untitled #458)",
"visual_prompt": "Studio Portrait. [Subject] wearing a short, stiff brown wig with bangs. Makeup: Heavy, caked powder, aggressive black eyeliner, overlined red lips showing teeth in a manic smile. Wardrobe: Oversized jacket with NEON YELLOW FUR lapels and intricate GEOMETRIC TRIBAL patterns on sleeves. Background: Smoky, wavy, gradient green abstract backdrop. Lighting: Hard, flat studio strobe (commercial look). Texture: Synthetic fur, caked foundation."
},
{
"frame_id": 2,
"archetype": "The Sweaty Centerfold (Ref: Untitled #93)",
"visual_prompt": "High-angle POV. [Subject] lying in messy DARK BLUE/BLACK bedding. Skin texture: Extremely OILY, SWEATY, and glistening (simulating humidity). Wig: Messy, matted blonde. Expression: Teary-eyed, distressed, looking up. Lighting: Strong TUNGSTEN (Orange/Warm) directional light casting deep shadows. Vibe: Claustrophobic, sticky."
},
{
"frame_id": 3,
"archetype": "The Grunge Punk (Ref: Untitled #129)",
"visual_prompt": "[Subject] crouching on a dirty floor. Wardrobe: Purple metallic fringed skirt spreading on floor, grunge striped top. Wig: Dark, choppy, unkempt bangs. Lighting: Theatrical Split-Gel lighting—Harsh RED rim light on hair/shoulder, Cold BLUE fill on face. Expression: Dissociated, blank stare. Atmosphere: Gritty, backstage clutter."
},
{
"frame_id": 4,
"archetype": "The Forensic Corpse (Ref: Untitled #153)",
"visual_prompt": "Macabre overhead shot. [Subject] lying supine on distinct MOSSY/DIRT ground. Skin Tone: UNNATURAL PALLOR (Grey/Blue undertones of death). Face Detail: Thick MUD smeared (not dotted) on cheeks and forehead. Eyes: Wide open, glassy, staring at nothing. Lighting: Flat, cold, clinical daylight (no shadows). Vibe: Uncanny valley, forensic photography."
},
{
"frame_id": 5,
"archetype": "The Old Master (Ref: Untitled #224)",
"visual_prompt": "History Portrait. [Subject] styled as a 'Bad' Caravaggio painting. Wardrobe: Cheaply draped red and gold fabrics (toga style). Props: Holding obviously plastic/fake grapes. Skin: Yellowish 'jaundice' tint typical of old varnish. Lighting: Deep Chiaroscuro (extreme shadow vs light). Vibe: Satirical museum piece, visible prosthetics."
},
{
"frame_id": 6,
"archetype": "The Noir Heroine (Ref: Untitled Film Still #21)",
"visual_prompt": "Black & White 35mm. [Subject] standing in a vintage city street canyon. Wardrobe: 1950s structured suit and small hat. Lighting: High-contrast natural sunlight creating sharp architectural shadows. Expression: Apprehensive, looking off-camera. Composition: Subject is small in frame, dominated by out-of-focus skyscrapers. Grain: Heavy film grain."
}
]
}
}David Lachapelle style images
Original:
David Lachapelle contact sheet images:
