Image to JSON prompt: structured prompts you can version and reuse.
A JSON prompt is the same instruction as a written one, with the parts named. Instead of a sentence where subject, lighting and lens are tangled into prose, each is a field you can read, change or hold constant. Geno produces one from any reference image, alongside the plain-text and assembled forms.
Why a structured prompt beats a sentence
Prose is fine for one image. It stops being fine the moment you need the second one to match, because there is no reliable way to change exactly one thing in a paragraph. Named fields give you that: hold the lighting and the lens, change the subject, and the two images still belong together.
It also makes a prompt reviewable. A field you can point at is a field a colleague can question, and a diff between two JSON prompts says precisely what changed — which a diff between two paragraphs does not.
What a JSON prompt from Geno looks like
Flat, and keyed by attribute. This is a real decode, beside the image it produced — the prompt on the right is what generated the picture on the left, unedited:
{
"primary_subject": "A person with sleek center-parted dark hair styled into intricate face-framing braids and a thick sculptural side ponytail wrapped in dark metallic-look cuffs, wearing a pale sage-white technical zip-up jacket with a high stand collar; face shows natural skin texture with visible freckles, a small symbol tattoo near the temple, a septum ring, and small dark decorative hair clips shaped like abstract wing or claw motifs.",
"secondary_subject": "Decorative hair accessories including dark ribbed cuffs binding the ponytail and small sculptural black clips near the ear, functioning as styling embellishments rather than separate subjects.",
"text_typography": "Bold black uppercase wordmark reading 'TRCKR' printed horizontally across the jacket's collar/chest area in a clean sans-serif athletic branding style, plus a small stylized glyph resembling a ram's-horn or omega-like symbol with a superscript '2' tattooed near the subject's temple.",
"color_palette": "Cool, desaturated palette dominated by pale ice-blue-white background and jacket, contrasted against deep blue-black hair and dark accessory hardware; skin tones read warm-neutral, creating a controlled two-tone tension.",
"composition": "Subject positioned right-of-center with the elaborate braided ponytail sweeping left to create a strong counterbalancing curve, generous negative space above the head and to the upper left, tight vertical framing that crops at the mid-chest allowing the high-collar jacket and hairstyle to dominate the lower two-thirds.",
"atmosphere": "The viewer senses a cool, controlled futurism—an air of quiet defiance and refined toughness radiating from the subject's steady gaze and architectural hairstyling.",
"story": "The image positions the subject as an emblem of next-generation athletic identity, blending technical sportswear with avant-garde beauty artistry to suggest innovation, individuality, and elevated street-luxury culture.",
"purchase_trigger": "The crisp branding on the jacket collar paired with the meticulous, editorial-level styling elevates the garment's desirability, implying exclusivity and cutting-edge design associated with the brand.",
"background_elements": "Seamless flat studio backdrop in pale grey-blue, evenly lit with no visible texture, gradient, or props, keeping full focus on the subject.",
"props": "Small dark sculptural hair clips resembling abstract wings or claws placed near the ear and hairline, plus ribbed dark cuffs binding sections of the ponytail; these function as styling props reinforcing the futuristic aesthetic.",
"arrangement": "Hair elements are asymmetrically arranged with braids swept to one side and the ponytail curving in a large loop to the opposite side, creating deliberate visual imbalance that draws the eye through the composition.",
"textures": "Glossy, tightly coiled braid texture in the ponytail; smooth slicked-back hair at the crown; matte technical nylon fabric on the jacket with visible zipper teeth and stitched seams; soft natural skin texture with fine freckling.",
"subject_pose": "Shoulders angled away from camera while the head turns back toward the lens, chin slightly lowered, creating a poised, introspective posture.",
"expression_gaze": "Neutral, slightly downward and off-axis gaze that avoids direct eye contact with the camera, giving a detached, editorial coolness.",
"wardrobe_styling": "Pale sage-white technical zip-up jacket with a tall structured collar, visible metal zipper pull, and bold branded wordmark across the chest/collar area; minimal accessories beyond the sculptural hair clips and a small septum ring.",
"hair_makeup": "Sleek center-parted hair with wet-look styling, intricate face-framing braids on one side, a thick wrapped ponytail on the other secured with ribbed dark cuffs; makeup is minimal, emphasizing natural skin texture and a soft matte finish.",
"skin_rendering": "Skin retains visible texture including freckles and fine tattoo linework, rendered with subtle sheen and soft even tonality, retouched for smoothness while avoiding a fully airbrushed look.",
"subject_action": "Static posed portrait, staged for a fashion/beauty campaign rather than candid, with the twist of the torso implying a paused mid-turn moment.",
"visual_style": "Clean high-fashion editorial with a futuristic/sci-fi styling edge, glossy retouched finish typical of luxury sportswear campaigns, controlled studio precision merged with avant-garde beauty styling.",
"camera_distance": "Close-up to medium close-up, framing from mid-chest to top of head, filling most of the vertical frame with the subject.",
"camera_angle": "Three-quarter angle with the face turned toward camera while the body is angled away, head slightly tilted down creating a contemplative, guarded look.",
"camera_height": "Eye level to very slightly above, aligned roughly with the subject's eyeline.",
"lens_specs": "Apparent short telephoto (85-100mm equivalent), minimal distortion on facial features, subtle background compression, clean edge sharpness with no visible vignette falloff.",
"depth_of_field": "Deep enough focus to keep face, braids, hair ornaments, and jacket collar all crisply rendered; background remains a smooth flat gradient rather than blurred bokeh, suggesting a wide aperture on a seamless backdrop rather than shallow subject-background separation.",
"film_sensor_look": "Clean high-resolution digital capture, noise-free, with a slightly cool color science; skin and hair textures rendered with hyper-detailed clarity suggesting heavy post-production sharpening.",
"post_processing": "Heavy but precise retouching that preserves skin texture (freckles, pores) while smoothing tonal transitions; cool-toned color grade with lifted shadows, contrast pushed on the black hair and jacket text for graphic punch, subtle skin dodge-and-burn for sculptural facial contouring.",
"depth_layers": "Flat seamless studio background sits behind the subject with no midground; foreground is entirely occupied by the subject's head, hair, and shoulders providing all depth cues through overlapping hair strands and jacket folds.",
"lighting_style": "Soft high-key studio lighting with an even, shadowless quality typical of beauty/fashion campaign photography.",
"key_light": "Large soft key light positioned front-left at roughly 30-40 degrees above eye level, producing gentle modeling on the face without harsh shadow lines.",
"lighting_quality": "Very soft and diffused, likely from a large octabox or scrim, wrapping smoothly around facial contours with minimal specular hotspots.",
"color_temperature": "Cool neutral, approximately 5600-6500K, giving the skin and background a slightly blue-grey cast consistent with the overall icy palette.",
"shadow_behavior": "Minimal, soft-edged shadows under the chin and along the jaw; shadows are barely perceptible and do not act as strong compositional elements, reinforcing the clean high-key look.",
"special_effects": "Subtle catch lights visible in the eyes; faint rim highlight along the hair strands and jacket zipper adding dimensional separation from the background.",
"aspect_ratio": "1:1",
"Hair Sculpting Technique": "Style the ponytail as a thick, uniformly coiled rope-braid bound at intervals with ribbed dark cuff segments, mimicking a mechanical or armor-like structure rather than a natural braid.",
"Facial Tattoo Placement": "Place a small stylized glyph/symbol tattoo with a superscript numeral near the outer temple/cheekbone area, rendered in fine thin linework matching natural skin tone contrast.",
"Ear Ornament Detail": "Attach small sculptural black clips shaped like abstract wings or claws directly onto sections of hair near the ear, functioning as jewelry-like styling accents integrated into the hairstyle.",
"Collar Branding Emphasis": "Render the jacket's stand collar oversized and structured enough to display bold horizontal wordmark branding clearly across the front, positioned just below the chin line."
}Thirty-eight fields, which is what a real photographic decode looks like rather than a sample written for a web page. Thirty-four of the keys are attributes from the taxonomy; the last four are custom attributes — things the model judged worth naming for this specific image, which is why they read as labels rather than ids. Nothing here is trimmed.
The full taxonomy spans 96 fields across four families — photographic, 3D render, illustration and graphic — and an image is only ever analysed against its own family’s slice, so a shot like this returns around 39, not 96. You choose which of them travel into the prompt, and dropping a field is how you tell the model you do not care about it.
Varying one field at a time
Because the fields are named, a set of images becomes a set of edits rather than a set of rewrites. Here is the same shoot in a different colourway — a burnt-orange snowboard shell instead of the pale technical jacket, under the same brand. It was generated from the prompt below with the first image attached as a reference, which is what holds the face steady while the clothes change.
{
"primary_subject": "A person with sleek center-parted dark hair styled into intricate face-framing braids and a thick sculptural side ponytail wrapped in dark metallic-look cuffs, wearing a burnt-orange technical snowboard shell with a high storm collar; face shows natural skin texture with visible freckles, a small symbol tattoo near the temple, a septum ring, and small dark decorative hair clips shaped like abstract wing or claw motifs.",
"secondary_subject": "Decorative hair accessories including dark ribbed cuffs binding the ponytail and small sculptural black clips near the ear, functioning as styling embellishments rather than separate subjects.",
"text_typography": "Bold black uppercase wordmark reading 'SNWR' printed horizontally across the shell's collar/chest area in a clean sans-serif athletic branding style, plus a small stylized glyph resembling a ram's-horn or omega-like symbol with a superscript '2' tattooed near the subject's temple.",
"color_palette": "Warm-cool complementary palette: a saturated burnt-orange shell set against the same pale ice-blue-white backdrop, with deep blue-black hair holding the cool end; skin tones read warm-neutral, and the orange-against-ice opposition replaces the original two-tone tension with a direct complementary contrast.",
"composition": "Subject positioned right-of-center with the elaborate braided ponytail sweeping left to create a strong counterbalancing curve, generous negative space above the head and to the upper left, tight vertical framing that crops at the mid-chest allowing the high-collar jacket and hairstyle to dominate the lower two-thirds.",
"atmosphere": "The viewer senses a cool, controlled futurism—an air of quiet defiance and refined toughness radiating from the subject's steady gaze and architectural hairstyling.",
"story": "The image positions the subject as an emblem of next-generation athletic identity, blending technical sportswear with avant-garde beauty artistry to suggest innovation, individuality, and elevated street-luxury culture.",
"purchase_trigger": "The crisp branding on the shell's storm collar paired with the meticulous, editorial-level styling elevates the garment's desirability, implying exclusivity and cutting-edge design associated with the brand.",
"background_elements": "Seamless flat studio backdrop in pale grey-blue, evenly lit with no visible texture, gradient, or props, keeping full focus on the subject.",
"props": "Small dark sculptural hair clips resembling abstract wings or claws placed near the ear and hairline, plus ribbed dark cuffs binding sections of the ponytail; these function as styling props reinforcing the futuristic aesthetic.",
"arrangement": "Hair elements are asymmetrically arranged with braids swept to one side and the ponytail curving in a large loop to the opposite side, creating deliberate visual imbalance that draws the eye through the composition.",
"textures": "Glossy, tightly coiled braid texture in the ponytail; smooth slicked-back hair at the crown; matte weatherproof ripstop shell fabric with visible taped seams and chunky zipper teeth; soft natural skin texture with fine freckling.",
"subject_pose": "Shoulders angled away from camera while the head turns back toward the lens, chin slightly lowered, creating a poised, introspective posture.",
"expression_gaze": "Neutral, slightly downward and off-axis gaze that avoids direct eye contact with the camera, giving a detached, editorial coolness.",
"wardrobe_styling": "Burnt-orange technical snowboard shell with a tall structured storm collar, exposed water-resistant chest zip with a chunky metal pull, articulated shoulder seams and a bold branded wordmark across the chest/collar area; minimal accessories beyond the sculptural hair clips and a small septum ring.",
"hair_makeup": "Sleek center-parted hair with wet-look styling, intricate face-framing braids on one side, a thick wrapped ponytail on the other secured with ribbed dark cuffs; makeup is minimal, emphasizing natural skin texture and a soft matte finish.",
"skin_rendering": "Skin retains visible texture including freckles and fine tattoo linework, rendered with subtle sheen and soft even tonality, retouched for smoothness while avoiding a fully airbrushed look.",
"subject_action": "Static posed portrait, staged for a fashion/beauty campaign rather than candid, with the twist of the torso implying a paused mid-turn moment.",
"visual_style": "Clean high-fashion editorial with a futuristic/sci-fi styling edge, glossy retouched finish typical of luxury sportswear campaigns, controlled studio precision merged with avant-garde beauty styling.",
"camera_distance": "Close-up to medium close-up, framing from mid-chest to top of head, filling most of the vertical frame with the subject.",
"camera_angle": "Three-quarter angle with the face turned toward camera while the body is angled away, head slightly tilted down creating a contemplative, guarded look.",
"camera_height": "Eye level to very slightly above, aligned roughly with the subject's eyeline.",
"lens_specs": "Apparent short telephoto (85-100mm equivalent), minimal distortion on facial features, subtle background compression, clean edge sharpness with no visible vignette falloff.",
"depth_of_field": "Deep enough focus to keep face, braids, hair ornaments, and jacket collar all crisply rendered; background remains a smooth flat gradient rather than blurred bokeh, suggesting a wide aperture on a seamless backdrop rather than shallow subject-background separation.",
"film_sensor_look": "Clean high-resolution digital capture, noise-free, with a slightly cool color science; skin and hair textures rendered with hyper-detailed clarity suggesting heavy post-production sharpening.",
"post_processing": "Heavy but precise retouching that preserves skin texture (freckles, pores) while smoothing tonal transitions; cool-toned color grade with lifted shadows, contrast pushed on the black hair and jacket text for graphic punch, subtle skin dodge-and-burn for sculptural facial contouring.",
"depth_layers": "Flat seamless studio background sits behind the subject with no midground; foreground is entirely occupied by the subject's head, hair, and shoulders providing all depth cues through overlapping hair strands and jacket folds.",
"lighting_style": "Soft high-key studio lighting with an even, shadowless quality typical of beauty/fashion campaign photography.",
"key_light": "Large soft key light positioned front-left at roughly 30-40 degrees above eye level, producing gentle modeling on the face without harsh shadow lines.",
"lighting_quality": "Very soft and diffused, likely from a large octabox or scrim, wrapping smoothly around facial contours with minimal specular hotspots.",
"color_temperature": "Cool neutral, approximately 5600-6500K, giving the skin and background a slightly blue-grey cast consistent with the overall icy palette.",
"shadow_behavior": "Minimal, soft-edged shadows under the chin and along the jaw; shadows are barely perceptible and do not act as strong compositional elements, reinforcing the clean high-key look.",
"special_effects": "Subtle catch lights visible in the eyes; faint rim highlight along the hair strands and the shell zipper adding dimensional separation from the background.",
"aspect_ratio": "1:1",
"Hair Sculpting Technique": "Style the ponytail as a thick, uniformly coiled rope-braid bound at intervals with ribbed dark cuff segments, mimicking a mechanical or armor-like structure rather than a natural braid.",
"Facial Tattoo Placement": "Place a small stylized glyph/symbol tattoo with a superscript numeral near the outer temple/cheekbone area, rendered in fine thin linework matching natural skin tone contrast.",
"Ear Ornament Detail": "Attach small sculptural black clips shaped like abstract wings or claws directly onto sections of hair near the ear, functioning as jewelry-like styling accents integrated into the hairstyle.",
"Collar Branding Emphasis": "Render the shell's storm collar oversized and structured enough to display bold horizontal wordmark branding clearly across the front, positioned just below the chin line."
}Eight of the thirty-eight fields differ. Two are the edit anyone sets out to make: wardrobe_styling and color_palette.
The other six are the part that catches people out. They do not describe the change, they merely mention the garment — primary_subject, text_typography, textures, purchase_trigger, special_effects and Collar Branding Emphasis. Leave them as they were and the prompt contradicts itself, asking for a pale sage jacket in one field and a burnt-orange shell in another, and the model splits the difference. This is the real skill in prompt editing, and it is only visible because the fields are separate: in a paragraph, those six mentions are just words you did not notice.
Everything else is held byte-identical, and that is what keeps it the same woman rather than a new one who happens to rhyme. Untouched: hair_makeup, skin_rendering, expression_gaze, subject_pose, camera_angle, camera_distance, camera_height, lens_specs, lighting_style, key_light, color_temperature, Hair Sculpting Technique, Facial Tattoo Placement and Ear Ornament Detail. Change expression_gaze or lens_specs here and you would get a different person in a different shot; the point of a variation is deciding what not to touch.
Looping your own script over one field is what this buys you, not a set of hand-written prompts you hope stay consistent.
Getting JSON out of Geno
Every analysis carries three forms and you switch between them on the result: full text for pasting somewhere that wants a sentence, assembled for a tuned single prompt, and JSON for everything above. Nothing is regenerated when you switch — it is the same decode, presented three ways.
If you want the broader picture first, the image to prompt page covers what the analysis returns and how the attributes are organised.
Try it on your own reference
Free trial credits on every account — sign in with Google, no card. What a credit buys is on the pricing page.