GPT Image 2.5 vs Nano Banana 2.1: 13 Use Cases Compared

Which image model fits your next project? Compare GPT Image 2.5 and Nano Banana 2.1 across ecommerce images, thumbnails, infographics, academic diagrams and more.

A product listing, a YouTube thumbnail and a scientific diagram need different things from an image model. A listing must preserve the product; a thumbnail must keep its title readable; a method diagram must connect the right boxes.

We gave both models the same brief for 13 use cases and compared their actual outputs. For five text-heavy tasks, we also generated separate English, Japanese and Korean versions. The main examples below use English copy or contain no headline text; the localized comparisons have their own Japanese and Korean copy.

The recommendations describe which model I would try first for these briefs. Each section shows the two independently generated results and explains the trade-off.

Try the examples on Felo: GPT Image 2.5, Nano Banana 2.1, or the AI Image tool.

Which model should you try first?

Use case	First choice in these samples
White-background product listings	Both; choose by finish
Product detail graphics	GPT for large callouts
Lifestyle product advertising	GPT for an advertising hero
Product concepts without a reference	Choose the intended object first
YouTube and tutorial thumbnails	GPT for bold title emphasis
Packing checklists and knowledge cards	GPT for large mobile type; NB for airy cards
Event posters	NB for graphic posters; GPT for venue mood
Numerical infographics	NB for this bar-chart brief
Process diagrams	GPT for training; NB for report figures
Scientific concept explanations	GPT for paper-style notation; NB for classroom slides
Research method and model-architecture diagrams	GPT for this seven-node framework
Story illustrations and concept art	GPT for rich texture; NB for a restrained scene
Interior-design concepts	GPT for atmosphere; NB for a simpler direction

The shared product reference

The first three ecommerce tests use the same fictional example product, including its left handle, ivory drinking lid and textured blue body.

1. White-background product listings

Shared brief: Keep the same blue mug, handle, drinking lid and proportions, remove the stone block, and place the complete product on white.

GPT Image 2.5	Nano Banana 2.1

Both models retained the handle and lid instead of turning the reference into a different travel cup. GPT Image 2.5 emphasizes texture and droplets; Nano Banana 2.1 gives the surface and contact shadow a softer finish. Neither produced an obvious overall winner for this one product.

When I would use it: Start with GPT if surface detail matters most; start with Nano Banana 2.1 if you prefer softer catalog lighting. For a real listing, compare the lid opening, handle curve, color and silhouette against the actual product before approving either output.

2. Product detail graphics

Shared brief: Use the same reference mug and three circular detail views, labeled Ceramic body, Ivory lid and Curved handle.

GPT Image 2.5	Nano Banana 2.1

Both English images preserve the requested headline and labels. GPT gives the texture and lid crops more room and uses larger feature labels. Nano Banana 2.1 makes smaller callouts around a centered product and retains more of the original stone setting. In several versions, GPT places a lid leader endpoint just above the product, so the line position still needs review.

When I would use it: For a mobile product-detail hero with prominent material close-ups, I would start with GPT. For a quieter supporting product page, Nano Banana 2.1 is a useful alternative. This choice is about emphasis and annotation cleanup, not evidence that one model preserves every SKU better.

Japanese version

GPT Image 2.5	Nano Banana 2.1

Both display カップの3つのディテール and the requested feature labels. GPT enlarges the texture crop, while Nano Banana 2.1 uses a more contained product panel. Check the leader-line endpoints rather than judging the headline alone.

Korean version

GPT Image 2.5	Nano Banana 2.1

Both display the three Korean feature labels. GPT uses wider, larger callouts; Nano Banana 2.1 uses a calmer layout. The curved-handle crop overlaps the main product more in the NB version.

3. Lifestyle product advertising

Shared brief: Place the same mug on an oak breakfast table beside one closed notebook, with a softly blurred kitchen and clear copy space on the right.

GPT Image 2.5	Nano Banana 2.1

GPT gives the mug stronger separation from the warm kitchen, with pronounced ceramic texture, wood grain and directional light. Nano Banana 2.1 keeps the scene calmer and lower in contrast. Both retain the main product structure and leave usable copy space.

When I would use it: Start with GPT for a brand banner or lifestyle ad where the product should lead the scene. Try Nano Banana 2.1 for a quieter everyday visual. Check product geometry, scale and shadows before adding campaign copy.

4. Product concepts without a reference

Shared brief: Describe a blue ceramic travel mug, ivory lid, limestone block and empty copy space, without supplying a product photograph.

GPT Image 2.5	Nano Banana 2.1

GPT produced a handled mug; Nano Banana 2.1 produced a slim handle-free cup. The prompt did not specify a handle, so these are two plausible interpretations of the same description. GPT emphasizes droplets and material texture, while Nano Banana 2.1 gives a simpler product silhouette.

When I would use it: Use GPT to explore a tactile advertising concept and Nano Banana 2.1 to explore a cleaner portable-cup direction. To depict an existing product, supply the same reference to both models; this text-only pair cannot establish product fidelity.

5. YouTube and tutorial thumbnails

Shared brief: Show hands photographing a night street with a phone, with BETTER PHONE PHOTOS and 3 COMPOSITION TIPS on the right.

GPT Image 2.5	Nano Banana 2.1

Both English thumbnails preserve the two requested text strings, while choosing different line breaks. GPT uses very heavy type and a compact secondary line. Nano Banana 2.1 uses narrower lettering, breaks the subtitle into two lines and gives the scene a more documentary appearance.

When I would use it: For a bold tutorial thumbnail, I would start with GPT. For a photography channel that prefers a less emphatic image, try Nano Banana 2.1. Check the final crop and title at the actual thumbnail size. These images were not tested for click-through rate; Japanese and Korean examples appear below.

Japanese version

GPT Image 2.5	Nano Banana 2.1

Both preserve スマホで映える写真 and 構図のコツ3選. GPT keeps the main headline on one line and separates the subtitle with a dark brush panel. NB splits the headline after スマホで and uses a stronger outline.

Korean version

GPT Image 2.5	Nano Banana 2.1

Both retain the recognizable Korean wording. GPT emphasizes the secondary line with a dark rounded panel; its word spacing looks tighter. NB separates the headline and subtitle more openly. Spacing should be reviewed in the final delivery.

6. Packing checklists and knowledge cards

Shared brief: Render six specified categories and item lists in a two-column, three-row packing checklist, with no added copy.

GPT Image 2.5	Nano Banana 2.1

The English versions both keep the six categories and their item lists. GPT gives the illustrations and labels more space within each card. Nano Banana 2.1 makes smaller icons and more whitespace; its passport icon also contains a tiny unrequested PASSPORT label. The Japanese and Korean versions do not repeat the extra English subtitle seen in the earlier Chinese experiment.

When I would use it: Choose GPT when large mobile-readable labels are your priority. Choose Nano Banana 2.1 when you prefer a more restrained card layout, and inspect icon labels as well as body copy. Do not transfer a copy error observed in one language into a claim about every language.

Japanese version

GPT Image 2.5	Nano Banana 2.1

Both show the six Japanese categories and their item lists without the unwanted English subtitle from the Chinese test. GPT favors larger illustrations; NB pairs smaller icons with compact text blocks.

Korean version

GPT Image 2.5	Nano Banana 2.1

Both show the six Korean categories and corresponding items. GPT keeps longer item lists on one line where possible; NB wraps the Bags item list across two lines. Both need final spacing and small-glyph review.

7. Event posters

Shared brief: Render NIGHT GARDEN and Botanical light show · Kyoto · 7–9 pm with luminous botanical shapes in green and amber.

GPT Image 2.5	Nano Banana 2.1

Both English posters retain the main event copy. GPT invents a garden venue with buildings and water. Nano Banana 2.1 frames the title with foliage and puts the subtitle near the bottom. Its Japanese and Korean versions group the event information more centrally.

When I would use it: Start with Nano Banana 2.1 for a graphic, typography-led event poster. Start with GPT for an atmospheric campaign visual. If a real venue matters, give both models a reference; the invented buildings should not imply a real location.

Japanese version

GPT Image 2.5	Nano Banana 2.1

Both retain ナイトガーデン and the location/time information. NB groups the title and subtitle centrally against foliage; GPT uses a smaller subtitle above a detailed garden scene. Check the time-range glyph when exact punctuation matters.

Korean version

GPT Image 2.5	Nano Banana 2.1

Both retain 나이트 가든 and the Korean location/time information. NB emphasizes the central text; GPT emphasizes an imagined garden venue. The distinction is primarily campaign style, not a proven difference in Korean accuracy.

8. Numerical infographics

Shared brief: Draw four bars on one zero baseline for 40%, 25%, 20% and 15%, with exact labels and a clear illustrative-data note.

GPT Image 2.5	Nano Banana 2.1

All six English, Japanese and Korean images preserve the displayed values and recognizable category labels. The bars themselves need another check. With the first bar normalized to 40, approximate pixel measurements of the English GPT bars give 25.6, 21.0 and 16.3; the Nano Banana 2.1 bars give 24.7, 20.1 and 15.6. Nano Banana 2.1 is closer to the requested proportions in these three language samples, although neither is numerically exact.

When I would use it: For this simple infographic draft, I would start with Nano Banana 2.1. For a chart whose geometry must be exact, render the final bars with a data charting tool. Correct percentages printed above bars do not establish that the bar heights are correct.

Japanese version

GPT Image 2.5	Nano Banana 2.1

The Japanese labels remain readable in both. NB adds a zero marker to the shared baseline. Its bar ratios are somewhat closer to the requested values, but the shortest bar is still a little high.

Korean version

GPT Image 2.5	Nano Banana 2.1

Both preserve the Korean labels and four percentages. Of these language samples, the Korean NB result is especially close to the specified bar-height ratios. This remains an approximate image, not a data-validated chart.

9. Process diagrams

Shared brief: Show Generate, Inspect and Refine in order, each with its specified supporting line, connected by arrows.

GPT Image 2.5	Nano Banana 2.1

Both outputs retain the labels, supporting text and three-step order. GPT uses thick blue arrows and rounded frames. Nano Banana 2.1 uses thinner connectors and quieter rectangular boxes. Both were requested in the same landscape format.

When I would use it: Choose GPT for a training slide where sequence should stand out immediately. Choose Nano Banana 2.1 for a supporting process figure in a report. This pair has no clear content-accuracy winner.

10. Scientific concept explanations

Shared brief: Explain a simplified diffusion process with four progressively noisy samples, followed by four denoising samples in reverse order.

GPT Image 2.5	Nano Banana 2.1

Both models preserve the sample counts, x0-to-x3 and x3-to-x0 ordering, and the requested arrow directions. GPT renders mathematical-looking subscripts and a stronger visible noise progression. Nano Banana 2.1 uses plain labels and a simpler sans-serif layout.

When I would use it: Start with GPT for a paper-style explanatory figure, or Nano Banana 2.1 for a straightforward classroom slide. These are conceptual illustrations, not experimental results. Check the actual paper notation and process before use. The underlying concept is discussed in Denoising Diffusion Probabilistic Models.

11. Research method and model-architecture diagrams

Shared brief: Draw a main image-processing path, a reference branch entering Feature fusion, and a dashed skip connection from Input image to Image decoder.

GPT Image 2.5	Nano Banana 2.1

Both models connect the main path, reference branch and skip connection to the requested boxes. GPT keeps the longer labels inside wider nodes and uses a clear orthogonal branch. Nano Banana 2.1 wraps labels into two lines, with its lower Reference labels close to the box edges.

When I would use it: I would start with GPT for this method diagram because less label-and-box adjustment is needed. Nano Banana 2.1 also gets the topology right. The framework is hypothetical; for a real paper, verify every arrow origin, destination and direction rather than treating a plausible diagram as evidence of the method.

12. Story illustrations and concept art

Shared brief: Put a tiny observatory and astronomer inside a transparent teacup on moss in a blue-hour forest.

GPT Image 2.5	Nano Banana 2.1

GPT creates a richly furnished miniature room and convincing glass reflections, but adds a second crescent moon. Nano Banana 2.1 gives the observatory a simpler structure and uses one moon above the cup. Neither perfectly implements every spatial instruction.

When I would use it: For a detailed fantasy concept, start with GPT and plan to correct the extra moon. For a quieter miniature illustration, start with Nano Banana 2.1. The more elaborate image is not automatically the more faithful one.

13. Interior-design concepts

Shared brief: Show a reading cafe with three round tables on the left, a continuous window-side bookcase, a rear counter and a clear center aisle.

GPT Image 2.5	Nano Banana 2.1

GPT emphasizes side light, oak textures and olive upholstery, and adds a rug, artwork and plants. Nano Banana 2.1 uses fewer decorations and a more direct view of the aisle and bookcase. Rear furniture overlaps, so exact table-and-chair counts require careful checking.

When I would use it: Start with GPT for an atmospheric client concept; start with Nano Banana 2.1 for a simpler visual direction. Neither perspective image establishes measured dimensions, circulation widths or a buildable plan.

Choosing for your own work

For a thumbnail or a product-detail hero, GPT Image 2.5 is a useful first attempt when you want prominent text and detail crops. For typography-led event posters, Nano Banana 2.1 provides a useful graphic direction. For the bar-chart brief, its geometry was closer to the supplied proportions across our three language samples.

For reference products, review the object before the setting. For scientific figures, review the relationships before the style. For localized marketing, review the actual target-language image: the extra English subtitle in an earlier Chinese checklist did not recur in our Japanese and Korean checklist samples.

Create your own images with these Felo tools

Choose a starting tool based on the examples above:

Your task	Recommended Felo tool
Product detail graphics, bold thumbnails, or research-method diagram drafts	Open GPT Image 2.5 on Felo
Graphic event posters, airy knowledge cards, or simple bar-chart drafts	Open Nano Banana 2.1 on Felo
Explore an image idea before choosing a model	AI Image

Copy a matching prompt from the appendix below. For a fair comparison, use the same prompt, reference image, and exact English, Japanese, or Korean copy in both model tools. Review product details, labels, and data before publishing.

Generation and review notes

The article contains 23 displayed pairs: 13 main cases and ten additional Japanese/Korean variants, for 46 comparison images. Each pair uses the same content prompt and, where relevant, the same reference. GPT ran through the requested ePhone GPT Image 2.5 Sunburst skill at high quality; Nano Banana 2.1 used its requested skill defaults. The quality settings and costs are not normalized. We retained the first successfully returned output for each displayed run and did not retouch the content.

Text observations are visual checks, not a native-speaker certification or repeated language benchmark. Small characters, Japanese time-range punctuation and Korean spacing still need delivery review. Bar proportions were measured approximately from pixels and are subject to antialiasing. No click-through rate, conversion rate or multi-run reliability was tested.

Read the exact prompts below.

Exact prompts used in the comparisons

Use the same prompt and reference image in both models. The product reference is shown above.

Ecommerce White

Create a square ecommerce catalog photograph using the supplied blue travel mug as the exact product reference. Preserve the cobalt-blue ceramic body, the curved handle on the left, the ivory drinking lid, the lid opening, the body proportions and surface texture. Remove the limestone block and the original environment. Show the complete mug centered on a pure white background, three-quarter view, with one subtle natural contact shadow beneath it. Product fills about 75% of the canvas height. No added objects, no text, no logo, no badge, no watermark. Do not redesign the product. Square 1:1.

Ecommerce Lifestyle

Create a realistic landscape lifestyle advertising photograph using the supplied blue travel mug as the exact product reference. Preserve its cobalt-blue ceramic body, curved handle, ivory drinking lid, and proportions. Place the complete mug on a light oak breakfast table next to one closed linen notebook in a quiet sunlit apartment kitchen. The mug is the sharp hero subject on the left third; the kitchen is softly out of focus. Warm morning light comes from the left. Leave the right third uncluttered as copy space. Natural believable scale and contact shadows. No people, no additional mugs, no text, no logo, no watermark. Landscape 16:9.

Academic Diffusion

Create a scientifically correct schematic for explaining a diffusion model in an academic presentation. Landscape 16:9, white background, precise thin lines, navy text, muted blue and gray palette, no decorative scenery. Exact title: 'Diffusion: noise and denoising'. Two clearly separated horizontal rows, each containing exactly four square image samples. The top row is labeled 'Forward process'. Its samples run left to right: a clean blue circle labeled 'x0', a lightly noisy version labeled 'x1', a very noisy version labeled 'x2', pure gray noise labeled 'x3'. Connect adjacent samples with three right-pointing arrows, each labeled 'Add noise'. The bottom row is labeled 'Reverse process'. Its samples run left to right: pure gray noise labeled 'x3', a very noisy blue circle labeled 'x2', a lightly noisy circle labeled 'x1', a clean blue circle labeled 'x0'. Connect adjacent samples with three right-pointing arrows, each labeled 'Denoise'. Preserve these exact labels, arrow directions, sample counts, and ordering. This is a conceptual explanation, not experimental results. No equations, no additional labels, no 3D, no watermark.

Interior Concept

Create a photorealistic interior design concept image of a small contemporary reading cafe, landscape 16:9. Wide view from the entrance at eye level, a clear continuous central aisle leading to the rear counter. Along the left wall, exactly three small round oak tables, each with exactly two chairs. Along the right wall, one continuous low oak bookcase beneath large windows. A compact coffee counter is centered on the rear wall. Warm white plaster, light oak, muted olive textiles, soft morning daylight from the right, believable furniture scale, coherent straight perspective, no people, no signage, no text, no extra tables, no watermark.

Academic Method

Create a precise academic method diagram for a hypothetical reference-guided image pipeline. White background, landscape 16:9, navy text, thin dark arrows, restrained pale blue rectangles, generous spacing, no decorative imagery. Exact title: 'Reference-guided image pipeline'. Use exactly seven boxes, each label appearing once. The main row runs left to right: 'Input image' -> 'Image encoder' -> 'Feature fusion' -> 'Image decoder' -> 'Output image'. A second row below the first two boxes contains 'Reference image' -> 'Reference encoder'. Draw one arrow from 'Reference encoder' upward and to the right into 'Feature fusion'. Draw one additional dashed skip-connection arrow from 'Input image' to 'Image decoder', routed above the main row and labeled exactly 'Skip connection'. All seven connectors must terminate on the specified boxes; do not connect the reference branch directly to the output or decoder. Keep every label readable and every box distinct. No extra boxes, no other labels, no graph statistics, no equations, no watermark. This is a hypothetical diagram, not an architecture attributed to a real paper.

Fantasy Scene

A cinematic editorial illustration of a tiny observatory built inside a clear glass teacup, resting on a moss-covered stone in a quiet forest at blue hour. A miniature astronomer in a rust-red coat looks through a brass telescope at a luminous crescent moon visible through the cup. Delicate reflections and condensation on the glass, rich tactile detail, restrained teal and amber palette, storybook realism, carefully composed, no text, no letters, landscape 4:3.

Product Concept

A premium studio product photograph of a matte cobalt-blue ceramic travel mug with an ivory lid, standing on a pale limestone block. Three-quarter view, soft morning window light from the upper left, subtle condensation, warm off-white background, generous negative space on the right, no logo, no text, no extra objects, realistic commercial photography, landscape 4:3.

Process Infographic

Create a clear horizontal infographic titled “A simple image review”. Show exactly three connected panels in this order, with these exact labels: “1. Generate”, “2. Inspect”, “3. Refine”. Add one small line under each: “Use one prompt”, “Check the details”, “Change one thing”. Flat editorial vector style, navy text, cobalt-blue connectors, white background, evenly spaced panels, no icons, no extra words, 16:9.

English, Japanese and Korean comparison prompts

ecommerce-detail-en

Language: English. Requested size: 1024x1280.

Specified text:

• Three details. One mug.
• Ceramic body
• Ivory lid
• Curved handle

Create a portrait ecommerce product detail graphic using the supplied blue travel mug as the exact product reference. Preserve its cobalt-blue ceramic body, left handle, ivory drinking lid, proportions, and texture. Put one full product view in the center, with three small circular detail crops around it: the ceramic body surface, the ivory lid, and the curved handle. Each detail crop must correspond to the same product. Add exactly this English headline: "Three details. One mug.". Add exactly these three labels next to their corresponding detail crops: "Ceramic body", "Ivory lid", "Curved handle". All text must be in English; do not add translations or another language. Use thin connector lines that point to the correct feature, generous spacing, a warm off-white background, navy text, refined commercial editorial layout. No invented performance claims, no numbers, no prices, no other words, no watermark. Portrait 4:5.

Reference: shared mug

data-infographic-en

Language: English. Requested size: 1536x864.

Specified text:

• Traffic sources
• Illustrative data, not real statistics
• Organic search
• Paid ads
• Direct
• Referrals
• 40%
• 25%
• 20%
• 15%

Design a precise flat vector infographic on a white background, landscape 16:9. Exact English title: "Traffic sources". Exact small subtitle: "Illustrative data, not real statistics". Draw exactly four vertical bars on one common zero baseline, left to right with these exact labels and values: "Organic search" 40%, "Paid ads" 25%, "Direct" 20%, "Referrals" 15%. Bar heights must be proportional to 40, 25, 20, 15: the first is tallest, the second is 62.5% of its height, the third is 50%, and the fourth is 37.5%. Put each value directly above its own bar and its category directly below. Use four restrained shades of blue, navy text, wide margins, readable typography. All text must be in English, with no translations or additional words. No perspective, no 3D, no icons, no extra charts, no watermark.

video-thumbnail-en

Language: English. Requested size: 1536x864.

Specified text:

• BETTER PHONE PHOTOS
• 3 COMPOSITION TIPS

Create a polished landscape 16:9 YouTube thumbnail for a English-language smartphone photography tutorial. On the left, a realistic close view of two hands holding one smartphone horizontally, its screen visibly framing the same neon-lit city street seen behind it. Use dramatic but believable teal and warm orange night lighting, a strong focal subject and crisp silhouettes. On the right, add exactly these two text strings in two distinct title groups: "BETTER PHONE PHOTOS" and "3 COMPOSITION TIPS". The first is larger; both must be readable at a thumbnail width of 320 pixels. Use bold high-contrast ivory typography with a subtle dark outline. Typography and line breaks must be natural for English, preserving every character exactly. All background signs must be abstract and unreadable. No other readable words, no translations, no camera brands, no logos, no comparison panels, no watermark. Keep all text inside the frame with generous margins.

event-poster-en

Language: English. Requested size: 1024x1024.

Specified text:

• NIGHT GARDEN
• Botanical light show · Kyoto · 7–9 pm

Design a square event poster with exactly this English headline: "NIGHT GARDEN". Add exactly this subtitle: "Botanical light show · Kyoto · 7–9 pm". Use luminous botanical silhouettes, deep forest green and warm amber, refined editorial typography, clear contrast, balanced margins, all text fully inside the canvas. Use natural English typography and line breaks without changing any characters. No other words, no translations, no logos, no watermark.

social-knowledge-en

Language: English. Requested size: 1024x1280.

Specified text:

• Weekend packing checklist
• Documents
• ID, tickets
• Clothes
• Jacket, spare clothes
• Electronics
• Phone, charger
• Toiletries
• Toothbrush, towel
• Bags
• Suitcase, organizers
• Essentials
• Tissues, umbrella

Design a readable portrait 4:5 English knowledge card for a travel packing checklist. Exact headline: "Weekend packing checklist". Use exactly six equal cards in a two-column by three-row grid. Left to right, top to bottom, each card has exactly this heading and one text line: "Documents" with "ID, tickets"; "Clothes" with "Jacket, spare clothes"; "Electronics" with "Phone, charger"; "Toiletries" with "Toothbrush, towel"; "Bags" with "Suitcase, organizers"; "Essentials" with "Tissues, umbrella". One simple relevant flat icon per card. Clean cream background, navy typography, restrained blue and orange accents, clear title hierarchy, comfortable whitespace, no overlapping elements. Use natural English typography and line breaks. Preserve all specified text and punctuation exactly. No extra copy, no translations, no other languages, no watermark.

ecommerce-detail-ja

Language: Japanese. Requested size: 1024x1280.

Specified text:

• カップの3つのディテール
• 陶器の本体
• アイボリーのふた
• カーブした持ち手

Create a portrait ecommerce product detail graphic using the supplied blue travel mug as the exact product reference. Preserve its cobalt-blue ceramic body, left handle, ivory drinking lid, proportions, and texture. Put one full product view in the center, with three small circular detail crops around it: the ceramic body surface, the ivory lid, and the curved handle. Each detail crop must correspond to the same product. Add exactly this Japanese headline: "カップの3つのディテール". Add exactly these three labels next to their corresponding detail crops: "陶器の本体", "アイボリーのふた", "カーブした持ち手". All text must be in Japanese; do not add translations or another language. Use thin connector lines that point to the correct feature, generous spacing, a warm off-white background, navy text, refined commercial editorial layout. No invented performance claims, no numbers, no prices, no other words, no watermark. Portrait 4:5.

Reference: shared mug

data-infographic-ja

Language: Japanese. Requested size: 1536x864.

Specified text:

• 流入元
• 説明用のデータです
• 自然検索
• 広告
• 直接アクセス
• 参照サイト
• 40%
• 25%
• 20%
• 15%

Design a precise flat vector infographic on a white background, landscape 16:9. Exact Japanese title: "流入元". Exact small subtitle: "説明用のデータです". Draw exactly four vertical bars on one common zero baseline, left to right with these exact labels and values: "自然検索" 40%, "広告" 25%, "直接アクセス" 20%, "参照サイト" 15%. Bar heights must be proportional to 40, 25, 20, 15: the first is tallest, the second is 62.5% of its height, the third is 50%, and the fourth is 37.5%. Put each value directly above its own bar and its category directly below. Use four restrained shades of blue, navy text, wide margins, readable typography. All text must be in Japanese, with no translations or additional words. No perspective, no 3D, no icons, no extra charts, no watermark.

video-thumbnail-ja

Language: Japanese. Requested size: 1536x864.

Specified text:

• スマホで映える写真
• 構図のコツ3選

Create a polished landscape 16:9 YouTube thumbnail for a Japanese-language smartphone photography tutorial. On the left, a realistic close view of two hands holding one smartphone horizontally, its screen visibly framing the same neon-lit city street seen behind it. Use dramatic but believable teal and warm orange night lighting, a strong focal subject and crisp silhouettes. On the right, add exactly these two text strings in two distinct title groups: "スマホで映える写真" and "構図のコツ3選". The first is larger; both must be readable at a thumbnail width of 320 pixels. Use bold high-contrast ivory typography with a subtle dark outline. Typography and line breaks must be natural for Japanese, preserving every character exactly. All background signs must be abstract and unreadable. No other readable words, no translations, no camera brands, no logos, no comparison panels, no watermark. Keep all text inside the frame with generous margins.

event-poster-ja

Language: Japanese. Requested size: 1024x1024.

Specified text:

• ナイトガーデン
• 植物のライトショー・京都・19〜21時

Design a square event poster with exactly this Japanese headline: "ナイトガーデン". Add exactly this subtitle: "植物のライトショー・京都・19〜21時". Use luminous botanical silhouettes, deep forest green and warm amber, refined editorial typography, clear contrast, balanced margins, all text fully inside the canvas. Use natural Japanese typography and line breaks without changing any characters. No other words, no translations, no logos, no watermark.

social-knowledge-ja

Language: Japanese. Requested size: 1024x1280.

Specified text:

• 週末旅行の持ち物リスト
• 書類
• 身分証、チケット
• 衣類
• 上着、着替え
• 電子機器
• スマホ、充電器
• 洗面用品
• 歯ブラシ、タオル
• 収納
• スーツケース、収納袋
• 日用品
• ティッシュ、傘

Design a readable portrait 4:5 Japanese knowledge card for a travel packing checklist. Exact headline: "週末旅行の持ち物リスト". Use exactly six equal cards in a two-column by three-row grid. Left to right, top to bottom, each card has exactly this heading and one text line: "書類" with "身分証、チケット"; "衣類" with "上着、着替え"; "電子機器" with "スマホ、充電器"; "洗面用品" with "歯ブラシ、タオル"; "収納" with "スーツケース、収納袋"; "日用品" with "ティッシュ、傘". One simple relevant flat icon per card. Clean cream background, navy typography, restrained blue and orange accents, clear title hierarchy, comfortable whitespace, no overlapping elements. Use natural Japanese typography and line breaks. Preserve all specified text and punctuation exactly. No extra copy, no translations, no other languages, no watermark.

ecommerce-detail-ko

Language: Korean. Requested size: 1024x1280.

Specified text:

• 컵의 세 가지 디테일
• 세라믹 바디
• 아이보리 뚜껑
• 곡선형 손잡이

Create a portrait ecommerce product detail graphic using the supplied blue travel mug as the exact product reference. Preserve its cobalt-blue ceramic body, left handle, ivory drinking lid, proportions, and texture. Put one full product view in the center, with three small circular detail crops around it: the ceramic body surface, the ivory lid, and the curved handle. Each detail crop must correspond to the same product. Add exactly this Korean headline: "컵의 세 가지 디테일". Add exactly these three labels next to their corresponding detail crops: "세라믹 바디", "아이보리 뚜껑", "곡선형 손잡이". All text must be in Korean; do not add translations or another language. Use thin connector lines that point to the correct feature, generous spacing, a warm off-white background, navy text, refined commercial editorial layout. No invented performance claims, no numbers, no prices, no other words, no watermark. Portrait 4:5.

Reference: shared mug

data-infographic-ko

Language: Korean. Requested size: 1536x864.

Specified text:

• 유입 경로
• 설명용 예시 데이터
• 자연 검색
• 유료 광고
• 직접 방문
• 추천 링크
• 40%
• 25%
• 20%
• 15%

Design a precise flat vector infographic on a white background, landscape 16:9. Exact Korean title: "유입 경로". Exact small subtitle: "설명용 예시 데이터". Draw exactly four vertical bars on one common zero baseline, left to right with these exact labels and values: "자연 검색" 40%, "유료 광고" 25%, "직접 방문" 20%, "추천 링크" 15%. Bar heights must be proportional to 40, 25, 20, 15: the first is tallest, the second is 62.5% of its height, the third is 50%, and the fourth is 37.5%. Put each value directly above its own bar and its category directly below. Use four restrained shades of blue, navy text, wide margins, readable typography. All text must be in Korean, with no translations or additional words. No perspective, no 3D, no icons, no extra charts, no watermark.

video-thumbnail-ko

Language: Korean. Requested size: 1536x864.

Specified text:

• 스마트폰 사진 잘 찍기
• 구도 팁 3가지

Create a polished landscape 16:9 YouTube thumbnail for a Korean-language smartphone photography tutorial. On the left, a realistic close view of two hands holding one smartphone horizontally, its screen visibly framing the same neon-lit city street seen behind it. Use dramatic but believable teal and warm orange night lighting, a strong focal subject and crisp silhouettes. On the right, add exactly these two text strings in two distinct title groups: "스마트폰 사진 잘 찍기" and "구도 팁 3가지". The first is larger; both must be readable at a thumbnail width of 320 pixels. Use bold high-contrast ivory typography with a subtle dark outline. Typography and line breaks must be natural for Korean, preserving every character exactly. All background signs must be abstract and unreadable. No other readable words, no translations, no camera brands, no logos, no comparison panels, no watermark. Keep all text inside the frame with generous margins.

event-poster-ko

Language: Korean. Requested size: 1024x1024.

Specified text:

• 나이트 가든
• 식물 라이트 쇼 · 교토 · 오후 7–9시

Design a square event poster with exactly this Korean headline: "나이트 가든". Add exactly this subtitle: "식물 라이트 쇼 · 교토 · 오후 7–9시". Use luminous botanical silhouettes, deep forest green and warm amber, refined editorial typography, clear contrast, balanced margins, all text fully inside the canvas. Use natural Korean typography and line breaks without changing any characters. No other words, no translations, no logos, no watermark.

social-knowledge-ko

Language: Korean. Requested size: 1024x1280.

Specified text:

• 주말 여행 준비물
• 서류
• 신분증, 승차권
• 의류
• 겉옷, 여벌 옷
• 전자기기
• 휴대폰, 충전기
• 세면용품
• 칫솔, 수건
• 수납
• 여행 가방, 정리 파우치
• 생활용품
• 티슈, 우산

Design a readable portrait 4:5 Korean knowledge card for a travel packing checklist. Exact headline: "주말 여행 준비물". Use exactly six equal cards in a two-column by three-row grid. Left to right, top to bottom, each card has exactly this heading and one text line: "서류" with "신분증, 승차권"; "의류" with "겉옷, 여벌 옷"; "전자기기" with "휴대폰, 충전기"; "세면용품" with "칫솔, 수건"; "수납" with "여행 가방, 정리 파우치"; "생활용품" with "티슈, 우산". One simple relevant flat icon per card. Clean cream background, navy typography, restrained blue and orange accents, clear title hierarchy, comfortable whitespace, no overlapping elements. Use natural Korean typography and line breaks. Preserve all specified text and punctuation exactly. No extra copy, no translations, no other languages, no watermark.