创作工具从文字描述或参考图片开始

用 AI 横幅生成器把想法变成图片

为产品推广、活动和社交频道生成横幅。默认使用 GPT Image 2.5 Flare,也可选择 GPT Image 2 或 Nano Banana,支持文字描述和参考图片。

需登录 · 新账号获赠 5 积分 · 每张图片消耗 2 积分

生成横幅图片

选择模型,写下创意或添加参考图。默认使用 GPT Image 2.5 Flare。每张图片消耗 2 积分;新账号获赠 5 积分。

高级设置16:9 · 1

可选:上传 PNG、JPG 或 WEBP 图片,大小不超过 3MB,用作生成参考。

1 张图片 · 共 2 点(每张 2 点)

试试这些灵感
生成结果

生成的图片会显示在这里,可预览并下载。

AI 生成案例

用具体简报生成的广告、活动图和频道封面。品牌、人物和活动均为虚构演示;下方附所用模型、导出尺寸和完整提示词。

AI 生成的 LUMA COAST 夏季广告:珊瑚色水杯、浅蓝背景及 Summer, on repeat 标题。
GPT Image 2.5 Flare

夏季新品广告

LUMA COAST · 保温水杯新品广告,结合商品摄影、主标题和购买提示。

仅填入提示词和模型,不会自动生成或扣费。

查看完整生成提示词

AI 生成 · 虚构演示

模型
GPT Image 2.5 Flare
导出
1024 × 1024 px
源图
1254 × 1254 px

缩放后导出为 WebP

Create one finished square 1:1 e-commerce launch advertisement for the fictional reusable bottle brand LUMA COAST. Target export: 1080 by 1080 pixels. Art direction: premium summer product photography meets confident editorial typography. A single beautifully designed coral-orange stainless steel insulated bottle with a rounded pale-cream lid stands on a warm limestone plinth in the right half. Crisp sunlight, a subtle swimming-pool reflection, a soft palm-leaf shadow, pale turquoise background, tactile materials, sophisticated restraint. Make the bottle physically believable and free of existing brand logos. Set these exact words, with flawless spelling and strong hierarchy: small brand at top left "LUMA COAST"; large dark-navy headline on two lines "Summer, on repeat."; small supporting line "Meet your everyday bottle."; a clean dark-navy rounded CTA at lower left "Shop the drop". Keep the type on quiet, high-contrast space, with generous margins. The product occupies about 45 percent of the canvas; the headline is readable at mobile thumbnail scale. No discounts, prices, badges, testimonials, extra copy, watermark, device frame, collage, or third-party brand. Deliver only the flat advertisement artwork, not a mockup of an advertisement.

提示词保留原始简报中的目标尺寸;实际导出尺寸见上方。重新生成的结果会有所不同。

AI 生成的 LumaDesk 上线横幅:Your projects. One clear view 标题及项目看板示意。
GPT Image 2.5 Flare

SaaS 产品上线

LumaDesk · 用清晰的产品主张和原创界面示意介绍项目协作软件。

仅填入提示词和模型,不会自动生成或扣费。

查看完整生成提示词

AI 生成 · 虚构演示

模型
GPT Image 2.5 Flare
导出
1536 × 864 px
源图
1672 × 940 px

缩放后导出为 WebP

Create one finished 16:9 SaaS product-launch banner for the fictional project workspace LumaDesk. Target export: 1600 by 900 pixels. The product helps small teams organize projects. Use a warm off-white background, deep ink typography, one vivid cobalt accent, precise spacing, and a small original geometric brand symbol. Design it like a polished software launch campaign, calm and useful rather than a futuristic AI poster. The left 46 percent contains exact readable text: small brand "LumaDesk"; dominant headline split naturally into three lines "Your projects. One clear view."; secondary line "Plan together. Move work forward."; a cobalt CTA button shape labeled "Meet LumaDesk". On the right, show a clean original desktop-app panel at a very slight angle, with exactly three clearly named columns: "Plan", "In progress", "Done". Populate them with a few simple abstract cards and progress shapes, not tiny illegible text. Use restrained soft shadows and one subtle colored tab. Keep all type within generous safe margins. No browser chrome, real software logos, fabricated customer logos, growth charts, numerical performance claims, awards, extra text or watermark. One flat finished banner, not a web-page screenshot or multi-panel presentation.

提示词保留原始简报中的目标尺寸;实际导出尺寸见上方。重新生成的结果会有所不同。

AI 生成的 FORM + FIELD 工作坊横幅:群青背景、米白文字和几何造型。
GPT Image 2.5 Flare

线上工作坊报名

FORM + FIELD · 用主题、时间和报名提示推广虚构的设计系统工作坊。

仅填入提示词和模型,不会自动生成或扣费。

查看完整生成提示词

AI 生成 · 虚构演示

模型
GPT Image 2.5 Flare
导出
1536 × 864 px
源图
1672 × 941 px

缩放后导出为 WebP

Create one finished 16:9 promotional banner for a fictional online design workshop by FORM + FIELD. Target export: 1600 by 900 pixels. Art direction: contemporary design-conference poster, rich ultramarine background, warm ivory type, a single acid-yellow accent, crisp Swiss-inspired grid and playful original modular shapes. Use a large sculptural composition of interlocking paper-like geometric forms on the right third, with realistic soft texture. Keep the left two thirds quiet enough for excellent typography. Render only this exact copy: small brand "FORM + FIELD"; largest headline "Design systems, made practical."; a clear secondary information line "Live workshop · Oct 15 · 10 AM PT"; a concise CTA in an ivory outlined or solid pill "Save your seat". Make the hierarchy distinctive and the date genuinely readable. Leave at least 6 percent margin around the composition. The result should communicate a practical professional learning event, not a music festival. No invented speakers or portraits, no partner logos, no fake attendance claims, no QR code, no website URL, no extra microtext, no watermark. Output the actual flat promotional artwork, not a photograph of a poster.

提示词保留原始简报中的目标尺寸;实际导出尺寸见上方。重新生成的结果会有所不同。

AI 生成的 Maya Chen LinkedIn 封面:产品设计师文字、米白背景与折纸造型。
GPT Image 2.5 Flare

设计师 LinkedIn 封面

Maya Chen · 展示虚构产品设计师姓名、职业和设计方向的专业封面。

仅填入提示词和模型,不会自动生成或扣费。

查看完整生成提示词

AI 生成 · 虚构演示

模型
GPT Image 2.5 Flare
导出
1584 × 396 px
源图
1672 × 941 px

裁切、缩放后导出为 WebP

Create one finished LinkedIn personal-profile cover for an entirely fictional product designer named Maya Chen. Final target is a very wide 4:1 banner, 1584 by 396 pixels. If you must produce a 16:9 source image, treat its central horizontal band as the finished banner and keep all text between 37 and 59 percent of the full image height, with decorative continuation above and below for later cropping. Use a quiet warm-ivory background, charcoal typography, muted terracotta and indigo abstract folded-paper forms on the far right, precise editorial spacing and understated sophistication. Keep the lower-left region empty for a platform profile-photo overlay. Do not include a portrait, avatar circle, social-platform logo, platform interface, or an actual LinkedIn screenshot. Place a compact, left-aligned typography block in the center-right, between 35 and 82 percent of canvas width. Exact text only: large "MAYA CHEN"; medium "Product designer"; smaller "Turning complex ideas into clear experiences". Every line must survive the final 4:1 center crop. Do not put text in the top or bottom decorative continuation. No contact details, employer logos, awards, sales CTA, extra text or watermark. Deliver one flat original cover artwork.

提示词保留原始简报中的目标尺寸;实际导出尺寸见上方。重新生成的结果会有所不同。

AI 生成的 MOSSBYTE Twitch 横幅:温暖森林小屋、米白频道文字和直播时间。
GPT Image 2.5 Flare

治愈游戏直播频道

MOSSBYTE · 森林风格频道横幅,包含频道名称、内容定位和直播时间。

仅填入提示词和模型,不会自动生成或扣费。

查看完整生成提示词

AI 生成 · 虚构演示

模型
GPT Image 2.5 Flare
导出
1200 × 480 px
源图
1672 × 941 px

裁切、缩放后导出为 WebP

Create one finished static Twitch profile banner for a fictional cozy-gaming streamer called MOSSBYTE. Final target: 1200 by 480 pixels, 5:2 aspect ratio. If generated in 16:9, ensure the essential design survives a centered 5:2 crop: all text should fall between 34 and 59 percent of source-image height and between 8 and 48 percent of its width. Original art direction: a charming hand-painted woodland game world at blue hour, deep forest green, muted lilac, amber window lights, tiny rounded moss-covered cottages, a winding path, warm fireflies, and a small original mushroom character with a satchel on the far right. Make it feel like an inviting indie-game community, without referencing any existing game, character or artist. Leave a calm deep-green area toward the left for readable cream typography. Exact text: bold friendly headline "MOSSBYTE"; smaller line "Cozy games. Good company."; smallest readable schedule "TUE + THU · 8 PM". Keep the decorative scene extending to the right and place no important text near the outer edges. Do not include an offline-screen label, chat window, webcam box, follower count, game logos, platform logo, QR code, watermark or other text. One finished flat channel banner.

提示词保留原始简报中的目标尺寸;实际导出尺寸见上方。重新生成的结果会有所不同。

AI 生成的 SMALL WORLD NOTES YouTube 封面:海岸村庄与居中的慢旅行频道文字。
GPT Image 2

慢旅行 YouTube 频道

SMALL WORLD NOTES · 海岸风景频道封面,将频道名和标语集中放在画布中央。

仅填入提示词和模型,不会自动生成或扣费。

查看完整生成提示词

AI 生成 · 虚构演示

模型
GPT Image 2
导出
2560 × 1440 px
源图
1672 × 941 px

缩放后导出为 WebP

Create one finished YouTube channel-art image for the fictional slow-travel channel SMALL WORLD NOTES. A 16:9 landscape canvas intended for export at 2560 by 1440 pixels. All essential text must fit inside a compact central horizontal strip: x=22 to 78 percent of image width, y=39 to 61 percent of image height. This is a real channel-art composition, not a video thumbnail. Do not put title text near the top or bottom. Art direction: cinematic but gentle travel editorial photography, an original sunlit Mediterranean-style coastal village seen from a hillside, terracotta roofs, small fishing boats, soft turquoise sea, warm morning haze and natural film texture. Avoid recognizable landmark replicas and do not use a real travel-company logo. Let the scenery extend across the outer canvas, but keep the central text area visually calm with a subtle natural darker tonal area for contrast. In the exact center, render the title "SMALL WORLD NOTES" in elegant, widely spaced warm-white uppercase typography; directly below, render "Slow travel. Local stories." in a smaller clean typeface. Keep both lines easily readable inside the specified central strip. No other text, dates, flags, icons, faces, subscriber counts, fake platform buttons, watermark, device mockup or visible safe-area guide. Deliver one flat channel-art image.

提示词保留原始简报中的目标尺寸;实际导出尺寸见上方。重新生成的结果会有所不同。

三个步骤开始创作

从描述开始,检查生成结果,再为发布平台准备尺寸。

  1. 1. 描述画面

    写下主题、颜色、构图和需要留白的位置。可选上传一张参考图。

  2. 2. 选择设置并生成

    选择模型与画面比例,默认使用 GPT Image 2.5 Flare。登录后,每张图片消耗 2 积分。

  3. 3. 预览并下载

    检查文字、细节与构图。下载图片,并根据目标平台的尺寸和安全区域调整后再发布。

从描述到横幅图片

选择适合创意的模型,用文字和参考图片表达品牌、主题与构图。

三种生成模型

默认 GPT Image 2.5 Flare,也可切换 GPT Image 2 或 Nano Banana,均支持文字和参考图。

图片参考

上传一张不超过 3MB 的参考图,配合文字描述引导生成。

五种画面比例

可选 16:9、1:1、9:16、4:3 或 3:4。发布前请检查目标平台的像素尺寸。

预览与下载

下载生成图片,或用平台工具裁切并导出 PNG/JPEG。Nano Banana 还可在生成前选择输出格式。

可复用的创意示例

浏览电商、产品发布、活动和频道封面的原创 AI 示例,一键填入提示词,再修改成自己的创意。

按次使用积分

新账号获赠 5 积分,每张图片消耗 2 积分。提交前可查看账户余额。

生成积分计划

所有计划使用同一生成器。每张图片消耗 2 积分,以下为各计划的月度积分。

基础版

$12.00/ 月

按月订阅生成积分。

包含:

  • 每月 1,800 积分
  • 每张图片消耗 2 积分
  • 文字提示词与可选参考图片
  • 五种画面比例
  • GPT Image 2.5 Flare、GPT Image 2、Nano Banana
  • 图片预览与下载

请登录后购买积分

专业版

$19.50/ 月

按月订阅生成积分。

包含:

  • 每月 9,600 积分
  • 每张图片消耗 2 积分
  • 文字提示词与可选参考图片
  • 五种画面比例
  • GPT Image 2.5 Flare、GPT Image 2、Nano Banana
  • 图片预览与下载

请登录后购买积分

Max

$80.00/ 月

按月订阅生成积分。

包含:

  • 每月 19,200 积分
  • 每张图片消耗 2 积分
  • 文字提示词与可选参考图片
  • 五种画面比例
  • GPT Image 2.5 Flare、GPT Image 2、Nano Banana
  • 图片预览与下载

请登录后购买积分

常见问题

常见问题

了解输入方式、积分和下载前需要检查的内容。

AIBanner 能生成什么?

AIBanner 根据文字描述生成横幅图片,可选添加一张参考图。您可以描述社交媒体封面、频道背景或广告视觉创意。

默认使用什么模型?能切换吗?

默认使用 GPT Image 2.5 Flare,可在生成器中切换 GPT Image 2 或 Nano Banana。三种模型都支持文字描述和一张参考图,每张图片均消耗 2 积分。

需要登录吗?有多少免费积分?

生成图片需要登录。新账号创建后获赠 5 积分,有效期一年。每张图片消耗 2 积分。

如何使用参考图片?

在生成器中上传 PNG、JPG 或 WEBP 图片,大小不超过 3MB,并用文字描述希望生成的画面。参考图用于引导生成,输出细节仍需检查。

支持哪些比例与输出格式?

支持 16:9、1:1、9:16、4:3 和 3:4。GPT 模型下载其返回的图片;Nano Banana 可请求 PNG 或 JPEG。平台工具还能将图片裁切并导出为 PNG/JPEG,发布前请检查尺寸和安全区域。

能自动读取网站或管理品牌套件吗?

当前生成器使用文字描述和可选参考图片。网站 URL 不会自动转成品牌素材,也没有用于保存字体、标志和配色的品牌套件。

如何改善生成结果?

在描述中明确主题、颜色、构图和文字留白。若要重新生成,请修改提示词或参考图;每次新生成都会消耗积分。发布前检查文字拼写、细节及构图。

开始你的下一张横幅图片

登录后开始生成。新账号获赠 5 积分,每张图片消耗 2 积分。

AI 横幅生成器:文字与参考图生成横幅 | AIBanner