Chatgpt Images 20 Qwenimage30 Nano Banana...
Ratio 9:16Image-to-Image
Creator Reference:@akinonnon_kan
Original Post

Copy this Nano Banana Pro reference-guided prompt to create character...

暑い日は風鈴とスイカですね✨ Chatgpt Images 2. 0で生成 スレッドでQwen-image-3.

Master PromptClick to select all

暑い日は風鈴とスイカですね✨ Chatgpt Images 2.0で生成 スレッドでQwen-image-3.0 / Nano Banana で生成したものも載せておきます。 プロンプトは以下↓(要・参照画像) ---------------- An unretouched documentary-style photograph of the person from the reference image, shot on a full-frame digital camera with an 85mm prime. Infer and faithfully preserve apparent age, gender presentation, ancestry, face, hair, skin tone and texture, body shape and lines, height impression, proportions, and every physical characteristic exclusively from the reference; never beautify, age-shift, reshape, or invent traits. This applies to the torso in full: chest and bust size, shape and position, shoulder width, rib cage, waist, hips and limb thickness all come from the reference exactly as they appear there. Do not enlarge, reduce, lift, round, or otherwise alter the chest, and do not slim, lengthen or idealize any part of the body. Use the softened neckline and the fitted shorts only when the reference unambiguously depicts an adult; otherwise use a camisole whose neckline still sits at the collarbone and a straight-cut mid-thigh short with the same colors and mood. In either case the shorts stay fully opaque with no underwear visible. Scene: the wooden engawa veranda of an old Japanese house in late July, seen from the garden. Weathered dark timber posts and a run of paper shoji screens fill the background, an open doorway showing a dim room beyond. A small glass wind chime with a paper strip hangs from the eaves into the upper-left corner of the frame, at about head height beside the subject. The rim of a white enamel tray with cut watermelon enters the lower-right corner on the boards. Summer greenery presses in at the left edge. The boards are worn unevenly, with real grain and scuffs. Pose: the subject sits on the edge of the raised veranda in clean side profile to the camera, the front of the body facing out across the garden. One knee is drawn up with that bare foot resting on the boards; the other leg extends down off the veranda. The subject holds a wedge of watermelon in one hand and brings it to the mouth from the side of the face, that elbow resting near the raised knee; the other arm reaches back and the palm presses flat on the boards behind the hip, so the torso stays upright and open and nothing crosses in front of the chest. The gaze goes out across the garden, away from the camera; the expression is languid and unposed, caught mid-bite rather than held for the shot. Outfit: a much-loved everyday camisole in sand-ivory cotton jersey, clearly old and washed many times: the knit has gone soft and slightly limp, the bound neckline has lost its original tension and now sits a little lower and looser than it was made to, and the thin straps have stretched with wear so they sit slack rather than taut, though both stay properly on the shoulders. The color has faded unevenly, the hem curls slightly, and the fabric carries the settled creases of a garment worn at home all summer. It remains fully opaque, unstructured and unpadded, with no shaping or support built into it, so it simply follows the reference body. Slate-gray cotton-linen shorts, equally well-worn and soft, cut close to the body with the hem ending high on the thigh, a relaxed drawstring left untied, and light creases from sitting. Hero visual: every vertical in the frame holds perfectly still — the timber posts, the shoji rails, the hanging cord of the wind chime — and the paper strip beneath the chime is the single element in the photograph caught mid-swing, hanging just off plumb. Camera and framing: a medium shot taken from the garden at roughly seated shoulder height, level with no tilt, positioned square to the subject's profile. The frame runs from just above the head down to mid-thigh: the whole face, both shoulders, the full torso, the waist, the seated hip and the top half of the thighs all read clearly and fill the frame. Crop below the knee; the lower legs and feet stay out of shot. Never position the camera behind the subject; the front of the torso and the profile of the face must both read. Never frame between the legs. Lighting: available light only, overcast-bright midday summer. Soft even daylight from the open garden, the veranda interior falling into deep natural shade behind the subject, cool green reflected fill on the skin from the grass, a slight warm return off the timber. No studio light, no rim light, no artificial fill. Photographic realism: real optical behavior throughout — moderate depth of field around f/2.8 with the background shoji and greenery gently soft, focus on the face, mild lens vignetting, faint chromatic fringing on the brightest highlights, natural sensor grain in the shadowed interior, slightly clipped specular highlights on the glass chime and the enamel tray. No skin smoothing, no airbrushing, no retouching of any kind. Fabric drapes under real gravity with genuine creases. Cotton jersey behaves like real worn knit: soft folds, no crispness, no stiff structure. Format: 3:4 vertical. A candid photograph, not a render or an illustration — no CGI look, no plastic skin, no HDR glow, no digital sharpening halos, no beauty filter.

Optimal Generation Parameters
Aspect Ratio
9:16
ModelNano Banana Pro
Aesthetic KeynotesMinimalist, 3D Rendered, Anime-Inspired, Moody

Test & Generate with GPT Image 2

Prompt tokens and APIMart GPT Image 2 are pre-loaded. Customize or click generate.

GPT Image 2 生成器

Copy this Nano Banana Pro reference-guided prompt to create character...
Nano Banana 2
Creative Blueprint & Recipe Breakdown

How to Master & Adapt This Visual

Professional prompts rely on modular tokens. Use this recipe breakdown to swap subjects, control studio lighting, and eliminate common AI artifacts.

Prompt Variables Blueprint (What to Customize First)

Iterate on one variable group at a time to maintain stylistic consistency:

1Change the subject block first: 暑い日は風鈴とスイカですね✨ Chatgpt Images 2.
2Adjust the mood through light and backdrop cues: Weathered dark timber posts and a run of paper shoji screens fill the background, an open doorway showing a dim room...
3Keep the finish constraints aligned while you test variants: Infer and faithfully preserve apparent age, gender presentation, ancestry, face, hair, skin tone and texture, body shape...

3 Style Recipes

Swap the subject, keep the frameRecipe #1

Replace 暑い日は風鈴とスイカですね✨ chatgpt images 2. with your own subject while leaving the camera and composition cues intact.

Retune the mood, not the structureRecipe #2

Keep the subject block, then rewrite weathered dark timber posts and a run of paper shoji screens fill the background, an... to move the image into a new lighting setup.

Push the restyle with a cleaner referenceRecipe #3

Start from a sharper or more contrasty source image, then edit finish descriptors only after subject preservation is stable.

Production Quality Guide: Best Use vs. Pitfalls

Ideal Use Cases (Where it Shines)
  • •Portrait studies, beauty concepts, and tightly framed editorial references.
  • •Reference-image workflows where identity, product shape, or silhouette should survive the restyle.
  • •Clean studio results where spacing, surfaces, and product separation need to read immediately.
  • ✓Start with a sharp reference image. This prompt is written to preserve source identity, silhouette, or product structure before stylization.
  • ✓Use 9:16 on the first run, then test alternate crops after the base composition is working.
  • ✓Lock the environment before fine styling changes: Weathered dark timber posts and a run of paper shoji screens fill the background, an open doorway showing a...
Pitfalls to Avoid (Common Mistakes)
  • !Ignoring the recommended 9:16 crop on the first run. The prompt's composition cues were written for that frame.
  • !Using a muddy reference image and expecting the prompt to repair identity, silhouette, or packaging structure on its own.
  • !Dropping the lighting or backdrop cues too early. Those clauses usually carry most of the mood and depth.
  • ✕Skip if: You need a wide environment shot or multiple subjects competing for attention in the same frame.
  • ✕Skip if: You need to begin with a radically different crop than 9:16 instead of honoring the existing composition hints.
  • ✕Skip if: Your reference image is low-resolution, badly lit, or cluttered enough to break subject preservation.

Technical Analysis & How This Prompt Works

暑い日は風鈴とスイカですね✨ Chatgpt Images 2. 0で生成 スレッドでQwen-image-3. Best in 9:16. Built for reference-image workflows.

Use a clean reference image with obvious subject edges and readable facial or product details before you start changing the styling. Keep the generator on 9:16 for the closest match. The mood is anchored by weathered dark timber posts and a run of paper shoji screens fill the background, an open doorway showing a...

Explore Related Curated Categories

More Visual Inspirations

Explore prompts sharing similar aesthetic tokens and composition

View all prompts →