Kling Image: The Essentials

Kling is Kuaishou's image family, and on Scenario it comes as four models across two generations. Kling V3 handles text to image and image to image at 1K or 2K, with rich cinematic color and strong editorial and illustration looks. Kling O3 is the Omni generation: it reaches native 4K, follows long prompts closely, and its image to image model can edit or combine up to 10 reference images in one prompt. All four support elements, a face-control feature that keeps up to three people consistent across new scenes.

Which Model Should I Use?

Starting from text, use V3 for speed and range at 2K and O3 when you need 4K or very literal adherence to a long prompt. Starting from one image you want to keep the layout of, use V3 Image to Image. When the result has to combine several images, use O3 Image to Image.

How to Use the Models

Writing prompts for Kling

Kling reads natural language, so write a full paragraph rather than a list of tags: the subject and its details, the setting, the style or medium, the lighting, the composition and the mood. Kling recommends putting exclusions in the main prompt as plain sentences ("No text, no logos.") rather than relying on a negative prompt. For edits, Kling's own formula works well: "Keep everything else unchanged, modify [element] to [desired effect]". With elements, start with "Keep @Element1's facial features" and then describe the hair, clothing, action, background, lighting and atmosphere.

Text to Image with Kling V3

V3 gives rich, cinematic color across photography, illustration and stylized 3D. Every example below shows the real prompt and settings that produced it.

Documentary travel photography.

An elderly spice merchant arranging pyramids of saffron, sumac, turmeric and dried rose petals at a narrow stall in a covered Moroccan souk, brass scoops and woven baskets around him. Shafts of afternoon light fall through a slatted wooden roof and catch drifting spice dust in the air. Documentary travel photography, 35mm lens, rich warm color, shallow depth of field. Calm, unhurried mood. No text, no signage, no watermark.

Settings: aspectRatio 16:9, resolution 1K.

Spice merchant in a Moroccan souk, 16:9 at 1K · Open on Scenario

Fashion editorial. Strong backlighting and material rendering on a full-length vertical frame.

A fashion editorial portrait of a tall model in an oversized coat made of layered translucent amber resin panels, standing on a black volcanic beach at dawn, wind pulling at the coat's hem. Low sun behind her turns the resin panels into glowing honey-colored light, waves foam around her bare ankles, and fine mist hangs over the sea. Medium-format editorial photography, full-length composition, cool blue shadows against warm highlights, serene and otherworldly. No text, no logos.

Settings: aspectRatio 9:16, resolution 2K.

Amber resin coat on a volcanic beach, 9:16 at 2K · Open on Scenario

Game environment concept art, four variations in one run. With numOutputs set to 4, V3 returns four takes on the same brief.

Concept art for a fantasy game environment: a monastery of white stone towers and red lacquered bridges built across a chain of floating rock islands above a sea of clouds, waterfalls spilling off the island edges into the mist below. Tiny robed monks cross the bridges, prayer flags snap in the wind, and a giant paper kite shaped like a carp flies between the towers. Painterly matte-painting style, epic sense of scale, golden afternoon light with long soft shadows. No text, no UI.

Settings: aspectRatio 21:9, resolution 2K, numOutputs 4.

Floating monastery concept art, 21:9 at 2K · Open on Scenario

Anime key visual.

Anime key visual of a teenage courier on a battered red hover-scooter skimming low over a flooded city at sunrise, spraying a long arc of water behind her. She wears an oversized yellow raincoat, cargo shorts and a messenger bag stuffed with paper parcels, short black hair whipping in the wind, determined grin, leaning into the turn with both hands on the handlebars. Half-submerged apartment blocks with laundry lines, rooftop gardens and tangled power cables line the canal, and a flock of white herons lifts off ahead of her. Dramatic low angle tracking shot, strong motion in the water spray, crisp cel shading with clean line art, painted cumulus clouds glowing pink and gold, lens flare from the rising sun, vivid saturated palette like a theatrical film poster without any lettering. No text, no logos, no watermark, no subtitles.

Settings: aspectRatio 16:9, resolution 2K.

Courier over a flooded city, 16:9 at 2K · Open on Scenario

Short headline on a poster. One short word in quotation marks rendered cleanly and spelled correctly.

Vintage-inspired screen-printed film festival poster for a surf documentary. A lone surfer in silhouette paddles out toward an enormous curling wave that fills the upper two thirds of the frame, the wave rendered in bold flat layers of deep teal, turquoise and cream with stylized foam patterns. A large low sun in coral orange sits behind the wave. Visible halftone grain and slight paper texture, three to five flat ink colors, slightly misregistered edges for a hand-printed feel. The single headline word "TIDE" is set in large bold condensed sans-serif capital letters in cream across the bottom quarter, centered and perfectly spelled. No other text, no dates, no credits block, no logos, no watermark.

Settings: aspectRatio 2:3, resolution 2K.

Screen-printed surf poster, 2:3 at 2K · Open on Scenario

Clay stop-motion character.

3D clay render of a cheerful little mail-sorting robot in a cozy post office, styled like a handmade stop-motion puppet. The robot has a boxy body in pastel mint plasticine, a dome head with two big button eyes, stubby arms with three simple fingers holding a blank cream envelope, and a small rubber tank tread instead of feet. Behind it stands a wall of wooden pigeonhole cubbies stuffed with tiny colorful parcels and letters, with a brass bell on the counter and a potted cactus. Visible fingerprints and soft tool marks in the clay surfaces, slightly imperfect handmade proportions, warm soft key light from the upper left with gentle bounce light, shallow depth of field, miniature tabletop photography look. No writing on the envelopes or parcels, no text, no logos, no watermark.

Settings: aspectRatio 4:3, resolution 1K.

Mail-sorting robot in clay, 4:3 at 1K · Open on Scenario

Text to Image with Kling O3 (up to 4K)

O3 is the Omni image model. Set resolution to 4K for native high-resolution output, which holds fine detail such as dew drops, hair and distant architecture.

Landscape at native 4K.

A lone red cabin on stilts at the edge of a frozen fjord at blue hour, green aurora curtains rippling across the sky and reflecting in patches of clear black ice. A narrow line of footprints in the snow leads from the cabin's lit doorway down to a small wooden rowboat locked in the ice. Snow-covered mountains rise steeply on both sides. Ultra-detailed landscape photography, wide-angle lens, deep cold blues against the warm cabin window glow, vast quiet atmosphere. No text.

Settings: aspectRatio 16:9, resolution 4K.

Frozen fjord cabin under the aurora, 16:9 at 4K · Open on Scenario

Macro science image at 4K.

Extreme macro nature photograph of a small jumping spider perched on the curled tip of a fern frond covered in morning dew, shot for a science education feature. The spider faces the camera with its large glossy front eyes reflecting the sky, fuzzy iridescent hairs in copper and teal, its two front legs slightly raised, eight legs in total. Tiny spherical dew drops line the fern, each one refracting the green forest behind. The background dissolves into soft creamy bokeh of green and gold. Natural early morning backlight creating a delicate rim on the hairs, razor-thin depth of field with the eyes in critical focus, high detail, true color. Vertical composition with the spider in the upper third. No text, no logos, no watermark.

Settings: aspectRatio 9:16, resolution 4K.

Jumping spider on a dewy fern, 9:16 at 4K · Open on Scenario

Book cover illustration.

A children's book cover illustration: a small girl in a yellow raincoat rowing a teacup-shaped boat across a flooded city at night, a lantern hanging from her oar, curious goldfish the size of dogs swimming alongside her through the moonlit streets. Rooftops and chimneys poke out of the calm water, and a crescent moon reflects between them. Gouache painting style with visible brush texture, deep indigo and warm lantern orange palette, whimsical and brave mood. Leave empty space at the top for a title. No text.

Settings: aspectRatio 2:3, resolution 2K, numOutputs 2.

Gouache cover with space left for a title, 2:3 at 2K · Open on Scenario

Exhibition poster with a headline.

Modern exhibition poster for a botanical garden spring show. An explosive arrangement of oversized flowers fills the lower two thirds of the frame: coral peonies, deep blue delphiniums, striped tulips, white anemones with black centers and trailing sweet pea vines, all photographed in high detail against a flat saturated cobalt blue background. Petals overlap the side edges of the frame for a bold, abundant feel. Clean studio lighting with soft shadows giving the flowers a sculptural pop. The single headline word "BLOOM" is set in huge bold white sans-serif capital letters across the upper third, with the top of one peony overlapping the bottom of the letters for depth, perfectly spelled. No other text, no dates, no logos, no watermark.

Settings: aspectRatio 3:4, resolution 2K.

Botanical poster, 3:4 at 2K · Open on Scenario

Wide game key art.

Game splash art of a fleet of sand ships sailing across an endless golden dune sea at sunset, painted with the epic scale of a fantasy strategy game cover but with no interface. In the foreground a massive wooden flagship with patched crimson sails and huge bronze wheels instead of a keel crests a dune, sand pouring off its prow, a crew of tiny figures on the rigging. Behind it dozens of smaller ships spread out in formation, and on the horizon a colossal ruined statue half-buried in sand catches the last light. A swirling sandstorm wall approaches from the right, under a glowing amber sky with long violet shadows. Ultra-wide cinematic composition, painterly concept art with crisp detail in the foreground and atmospheric perspective in the distance. No UI, no title, no text, no logos, no watermark.

Settings: aspectRatio 21:9, resolution 4K.

Sand-ship armada, 21:9 at 4K · Open on Scenario

Consistent characters with elements

An element is one person (or object) defined by a frontal image plus one to three extra angles. Add up to three elements, then refer to them in the prompt as @Element1, @Element2 and @Element3. The face and identity carry into new clothes, poses, settings and even painted styles. The same field exists on all four models.

One element, V3 Text to Image. The frontal photo on the left (plus a three-quarter view) defines the person; the prompt places her in a new scene.

Keep @Element1's facial features. She has her silver hair pinned in a low bun, wearing a white beekeeper's jacket with the mesh veil pushed back over her shoulders and canvas gloves, holding up a dripping honeycomb frame and inspecting it with a proud smile. Background is a lavender field with stacked wooden beehives at golden hour, bees glinting in the backlight. Warm natural light, shallow depth of field, gentle and joyful atmosphere. No text, no logos.

Settings: aspectRatio 3:4, resolution 2K, one element (frontal plus one three-quarter view).

Element frontal image · Open on Scenario

Result, 3:4 at 2K · Open on Scenario

Two elements in one scene, V3 Text to Image. Each person keeps their own face and wardrobe.

Keep @Element1's facial features, short swept-back grey hair and salt-and-pepper beard, wearing his mustard-yellow cable-knit sweater with the sleeves pushed up and a leather work apron, seated at a cluttered clockmaker's workbench and holding an open brass pocket watch movement steady with tweezers. Keep @Element2's facial features, freckles and long copper-red braids, wearing her bottle-green raincoat, standing at his shoulder and leaning in close with a jeweler's loupe raised to her eye, studying the tiny gears with delight. The background is a narrow old repair shop lined with dozens of wall clocks, pendulums and wooden drawers of spare parts, rain streaking the shop window behind them. Warm pooled light from a green-shaded desk lamp on their faces and hands, cool blue daylight from the rainy window as a rim, cinematic medium two-shot at eye level, 50mm lens with shallow depth of field, a quiet, tender atmosphere of an apprentice's first lesson. Natural hands with five fingers each. No text, no clock numerals in focus, no logos, no watermark.

Settings: aspectRatio 16:9, resolution 2K, 2 elements.

@Element1 frontal · Open on Scenario

@Element2 frontal · Open on Scenario

Both elements together in a clockmaker's workshop, 16:9 at 2K · Open on Scenario

One element, O3 Text to Image.

Keep @Element1's facial features. He has close-cropped hair and a short beard, wearing a white chef's jacket with rolled sleeves and a dark apron, tossing vegetables in a flaming wok with one hand, flames leaping up and lighting his face. Background is a busy open restaurant kitchen at night with steel counters, hanging copper pans and cooks moving in soft focus. Warm orange firelight against cool steel tones, motion and energy, documentary food photography. No text, no logos.

Settings: aspectRatio 4:3, resolution 2K, one element (frontal plus profile view).

Element frontal image · Open on Scenario

Result, 4:3 at 2K · Open on Scenario

Editing one image with Kling V3 Image to Image

V3 Image to Image keeps the layout of the source very closely, which makes it a good fit for material swaps, relighting, season changes and restyles. Set aspectRatio to match the source when you want the same framing.

Material swap.

Keep everything else unchanged, modify the camera to be made entirely of chunky knitted yarn: a cream wool body, charcoal yarn leatherette panels, a lens barrel of tightly wound grey wool with a stitched glass-like lens, and tiny knitted dials on top. Keep the same position on the walnut desk, the same window light and the same camera angle, with soft fuzzy fiber texture visible on every surface. No text, no logos.

Settings: aspectRatio 4:3, resolution 1K.

Before · Open on Scenario

After: Camera remade in knitted yarn · Open on Scenario

Sketch to finished illustration.

Turn this pencil sketch into a finished full-color watercolor illustration for a children's storybook. Keep the exact composition, the oak tree, the treehouse shape with its porch railing, the round windows, the arched door and the rope ladder. Paint the oak with deep green foliage and a warm brown trunk, give the treehouse sky-blue shutters and a red tin roof, add golden lantern light in the windows, a fox and a small girl climbing the ladder, and a pale apricot evening sky. Soft paper texture, gentle color bleeds, cozy mood. No text.

Settings: aspectRatio 4:3, resolution 2K.

Before · Open on Scenario

After: Pencil sketch painted as a watercolor · Open on Scenario

Time of day and weather.

Keep the street, the yellow tram, the blue tiled facades, the hanging laundry and the camera angle unchanged, modify the time and weather to a rainy night. The cobblestones and tram rails become wet and reflect the tram's warm interior light and the glow of the iron streetlamps, the tiled facades shine with rain, windows above are lit in warm yellow, and soft fog hangs at the top of the hill. Cinematic night photography, teal and amber palette, fine rain streaks visible in the lamplight. No readable signage, no text.

Settings: aspectRatio 16:9, resolution 2K.

Before · Open on Scenario

After: Same street at night in the rain · Open on Scenario

Season change.

Keep everything else unchanged, modify the season from midsummer to deep winter. The same log cabin with the red tin roof and stone chimney stays in the same place, now with a thick layer of snow on the roof, a thin curl of smoke rising from the chimney and warm light glowing in the windows. The turquoise lake becomes partly frozen with pale blue ice and a dark open patch of water near the shore, the wooden dock and the rowboat are covered in snow, the wildflower meadow becomes smooth snowdrifts, the pine trees are frosted white, and the granite peaks are fully covered in snow. Cold clear late afternoon light with a soft pink sky. Same composition and camera position. No people, no text, no logos.

Settings: aspectRatio 16:9, resolution 2K.

Before · Open on Scenario

After: Summer lake cabin turned to winter · Open on Scenario

2D concept to 3D clay render.

Turn this flat 2D character concept into a 3D clay render while keeping the exact same character, pose and proportions. The small mushroom knight keeps the large red cap with cream spots and the ribbed gills underneath, the round pale face with two black dot eyes, the scratched silver breastplate and shoulder pauldrons with moss growing on the collar, the brown leather belt, the torn moss green tunic, the brown boots, the twig sword with a single green leaf raised in the right hand and the round acorn-cap shield on the left arm. Render it as a handmade plasticine stop-motion figure with soft fingerprint textures, rounded sculpted edges, matte clay surfaces, slightly metallic painted armor, on a seamless soft beige backdrop with warm studio key light and a soft ground shadow, shallow depth of field. No outlines, no text, no logos.

Settings: aspectRatio 3:4, resolution 2K.

Before · Open on Scenario

After: Flat character concept rebuilt as clay · Open on Scenario

Combining images with Kling O3 Image to Image

O3 Image to Image takes 1 to 10 images in referenceImages. Name each one in the prompt as @Image1, @Image2 and so on (or @Image when there is only one), and say exactly what to take from each: the product, the outfit, the face, the place or the style. aspectRatio auto picks the closest supported ratio to the main input.

Product into a scene (2 references).

Place the celadon teapot from @Image1 on the low wooden veranda of the moss garden in @Image2, beside a matching small cup with steam rising from it. Keep the teapot's exact shape, glaze color and bamboo handle from @Image1, and keep the garden, stone lantern and overcast light from @Image2. Add soft contact shadows under the teapot and a faint reflection of the garden on its glaze. Calm, contemplative product lifestyle photography. No text, no logos.

Settings: 2 reference images, aspectRatio auto, resolution 1K.

@Image1: product · Open on Scenario

@Image2: location · Open on Scenario

Result: the teapot placed on the veranda · Open on Scenario

Subject, garment and location (3 references, 4K).

The corgi from @Image1 wearing the cream cable-knit sweater from @Image2, sitting proudly on the wooden park bench from @Image3 among fallen orange maple leaves. Keep the corgi's exact markings and face from @Image1 and the sweater's knit pattern and red pom-pom collar from @Image2, fitted naturally around the dog's body. Soft late-afternoon autumn light, shallow depth of field, warm and cozy pet-lifestyle photography. No text, no logos.

Settings: 3 reference images, aspectRatio 3:2, resolution 4K.

@Image1: subject · Open on Scenario

@Image2: garment · Open on Scenario

Result with the bench from @Image3, 3:2 at 4K · Open on Scenario

Virtual try-on (person plus garment).

Virtual try-on. Keep @Image1's facial features, black chin-length bob, body proportions, standing pose, light blue straight-leg jeans, white canvas sneakers and the plain light gray studio backdrop exactly the same. Dress her in the cropped sage green quilted jacket from @Image2, worn open over her white t-shirt: same diamond quilting pattern, stand collar, dark green snap buttons, two flap chest pockets and ribbed cuffs, with realistic fit, natural fabric folds at the elbows and the cropped hem sitting at her waist above the jeans. Full body from head to shoes, even soft studio lighting, fashion e-commerce photography. No text, no logos, no labels.

Settings: 2 reference images, aspectRatio 2:3, resolution 2K.

@Image1: person · Open on Scenario

@Image2: garment · Open on Scenario

Result, 2:3 at 2K · Open on Scenario

Sketch to product render (4K).

Transform the sneaker sketch in @Image into a photorealistic product render, keeping the exact silhouette, the wavy layered midsole, the lug outsole and the panel lines from the sketch. Materials: off-white suede toe and heel panels, perforated forest-green leather side panels, a chunky gum rubber sole with visible tread, waxed cotton laces. Floating at a slight angle over a soft grey studio sweep with a gentle shadow below. Studio product photography, crisp detail on stitching and suede texture. No brand names, no logos, no text.

Settings: 1 reference image, aspectRatio 1:1, resolution 4K.

@Image: marker sketch · Open on Scenario

Result, 1:1 at 4K · Open on Scenario

Object removal on a single image.

Keep everything else in @Image unchanged, remove the red vintage scooter parked in front of the cafe entrance. Fill the area where it stood with what would naturally be behind it: the continuous gray stone sidewalk paving and curb, the lower dark green wooden panel of the cafe facade, the glass door and the bottom of the window, with the same morning sunlight and shadows. Keep the dark green awning, the round white marble bistro tables, the rattan chairs, the two potted olive trees in terracotta pots, the cobblestone street and the building facade exactly as they are. Clean seamless retouch with no smudges or ghosting. No people, no text, no logos.

Settings: 1 reference image, aspectRatio auto, resolution 2K.

Before · Open on Scenario

After · Open on Scenario

Five-image moodboard into one scene. Color swatches, a house, a textile, a car and a cactus combined into one 21:9 frame.

Turn this five-image moodboard into one cohesive cinematic scene. Use the color palette from @Image1 as the color grade of the whole image: burnt terracotta, dusty rose, sand beige, deep teal and olive. Use the sandstone adobe house from @Image2 as the setting, with its row of tall arched doorways, rounded parapets and the two tall fan palms. Use the teal and rust kilim from @Image3 as a rug laid on the ground in front of the arches, with its diamond patterns and cream fringe. Park the cream 1960s convertible from @Image4, with its tan interior, chrome bumpers and whitewall tires, in front of the house on the dusty ground. Use the lighting and mood from @Image5: a low golden hour sun with warm haze, long shadows and a tall columnar cactus silhouetted at the side of the frame against the glowing sky. Wide cinematic establishing shot, film photography look, no people. No text, no logos, no license plate.

Settings: 5 reference images, aspectRatio 21:9, resolution 2K.

Result from 5 references, 21:9 at 2K · Open on Scenario

Character, outfit and location (3 references, 4K).

Create a cinematic photograph combining three references. Take the man's identity from @Image1: keep his facial features, full white beard, short white hair, weathered tanned skin, blue eyes and smile lines. Dress him in the outfit from @Image2: the cream cable-knit fisherman sweater, the navy ribbed watch cap, the mustard yellow waxed canvas apron tied over the sweater and the dark green rubber boots. Place him in the location from @Image3: standing on the weathered gray wooden pier at dawn beside the stacked wire lobster traps and coiled orange ropes, with the blue and white fishing boat moored alongside and the misty pink and orange sky reflecting on the calm water behind him. He stands in a relaxed pose, full body, holding a coil of orange rope in both hands and looking out toward the water with a gentle smile. Warm soft dawn light on his face, realistic documentary photography, 35mm lens. No text on the boat, no logos.

Settings: 3 reference images, aspectRatio 16:9, resolution 4K.

Result, 16:9 at 4K · Open on Scenario

Parameters

The four models share most settings. Differences are called out in each entry.

prompt

Required on all four models, up to 2,500 characters. Write a full natural-language paragraph and put exclusions in it as sentences. Reference elements as @Element1 to @Element3 and, on O3 Image to Image, reference images as @Image1 to @Image10 (or @Image for a single image). See the poster examples for short quoted headlines.

negativePrompt

Kling V3 Text to Image only. Content to avoid. Kling itself recommends writing exclusions into the main prompt instead, which is what every example here does.

image

Kling V3 Image to Image only, required. The single reference image to transform. Minimum 300 by 300 pixels, aspect ratio between 0.4 and 2.5, up to 10 MB. See the before and after pairs above.

referenceImages

Kling O3 Image to Image only, required. 1 to 10 images, same size limits as above, referenced in the prompt by their order (@Image1 is the first one you add). See the try-on, product placement and moodboard examples.

aspectRatio

16:9 (default), 9:16, 1:1, 4:3, 3:4, 3:2, 2:3 or 21:9. O3 Image to Image also offers auto (its default), which picks the closest supported ratio to your main input. For edits that should keep the original framing, match the source ratio.

resolution

1K (default) or 2K on the V3 models. 1K, 2K or 4K on the O3 models; 4K is native (for example 5440 by 3072 at 16:9). The fjord, spider and sneaker examples were made at 4K.

numOutputs

1 to 9 images per run. Extra outputs are close variations of the same reading of the prompt, as in the four monastery variations, so use separate runs when you want clearly different results such as several colorways.

elements

Optional on all four models, up to 3. Each element needs a frontalImage (the main view) and 1 to 3 referenceImages from other angles. Reference them in the prompt as @Element1, @Element2, @Element3. See the beekeeper, workshop and chef examples.

Use Cases

• Game key art and concept art: wide 21:9 environments, isometric dioramas, pixel art scenes and character concepts, with O3 at 4K for marketing-ready splash art.

• Character marketing and storyboards: elements keep a cast recognizable across scenes, outfits and art styles, from photo to oil painting.

• E-commerce and product imagery: place a product into lifestyle scenes, swap backgrounds, mock up packaging and run virtual try-on with O3 Image to Image.

• Design iteration: turn sketches into finished renders or illustrations and explore materials, seasons and lighting on the same layout with V3 Image to Image.

• Publishing and education: book covers with title space, botanical posters and 4K macro science images.

• Editorial and campaigns: fashion, food and travel photography looks with cinematic color.

Tips for Better Results

1. Describe the structure of objects explicitly. O3 merged a "bottle with a nautilus-shaped stopper" into a single shell in our tests; spelling out what is attached to what avoids this.

2. Write exclusions into the prompt. Sentences such as "No text, no logos" were reliable across all four models; for O3, adding "plain" to surfaces like a tram front removed small invented emblems.

3. Use Kling's edit formula for single-image edits. "Keep everything else unchanged, modify X to Y" and listing the parts to keep produced the cleanest V3 Image to Image results.

4. Tell O3 what to take from each image. Name every reference (@Image1, @Image2) and the attribute you want from it: the face, the garment, the product shape, the location.

5. For style transfer, describe the target palette. A woodblock restyle only took hold once the prompt mapped the colors and line work explicitly instead of naming the style.

6. One run per variant. Several outputs in one run stay close to each other; run each colorway or concept separately.

7. Give elements a clear frontal image plus a real second angle. A plain background and even light on both views gave the strongest likeness.

8. Quote short headlines. One or two words in quotation marks rendered correctly on both generations; complex type layouts were simplified.

Known Limitations

• No Series mode on Scenario. Kling's own app can generate a connected image series; on Scenario, numOutputs gives independent variations. Use elements and a repeated style description to build a set.

• Secondary details can be dropped. Both generations sometimes simplify a requested action, pose or layout detail (for example a kneeling pose becomes standing). Put the most important detail early in the prompt and regenerate if needed.

• O3 Image to Image may reframe. It occasionally moves the camera even when asked to keep it. Use V3 Image to Image when the exact framing matters.

• Text layout is not precise. Short words render well, but font, overlap and placement instructions are often ignored. For dense typography, use a text-first model.

• Resolution limits. V3 tops out at 2K; only the O3 models offer 4K.

• Input limits. Reference and element images must be at least 300 by 300 pixels, between 0.4 and 2.5 in aspect ratio and under 10 MB.

• Cultural accuracy. Specific regional dishes, garments or objects can be approximated; check the result before using it as an authentic depiction.