速度
GPT Image 2.6
OpenAI 称生成速度比上一版本提升 50%。
AIGPTImage 现状
本站默认使用 2.6 Flare。单次最多可生成 4 张,具体取决于你的套餐。
OpenAI 发布 · 2026 年 11 月 4 日
OpenAI 推出了出图更快的新版本,还加入了动图输出和新的开发者控制项。下方生成器现已运行 GPT Image 2.6,这一页说明哪些新功能在本站可用,哪些目前只在 OpenAI 自己的产品里提供。
通过文字或参考照片创建图片。登录后即可生成图片,并在账户历史中保存和查看结果。
尽量具体,让 AI 更好理解你的创作方向
参考图可以帮助匹配你期望的结果
登录后可解锁私密生成历史和 20 个初始积分。
点击下方任意模板即可自动填入提示词
Ultra-detailed 3D plush cartoon character, inspired by classic animated TV style, standing in center frame, full body visible. Character designed as a soft toy with realistic fluffy fur texture, smooth rounded shapes, vibrant saturated colors, big expressive eyes, cute proportions, Using the image of [Doraemon's] character, [Nobita carries Doraemon on his back]. High-resolution Pixar-style 3D rendering, soft studio lighting, subtle rim light, clean gradient background (blue tone), cinematic lighting, depth of field, highly detailed fur strands, soft shadows under feet, glossy nose, toy-like material, ultra sharp 8K render. Centered composition, minimal background, character facing camera, symmetrical pose, adorable expression.
这里有 5 段循环动图,以及同一角色的 3 次编辑,全部由 Images 2.6 生成。为了让页面加载更快,原始 GIF 已转成体积更小的格式。
这些动图由 ChatGPT 中的 GPT Image 2.6 生成,提示词见下方。本页的生成器也可以生成 GIF。
在 ChatGPT 中想得到动图,提示词开头要写 “Create a GIF”;在本站生成器里,把输出格式选成 GIF 即可。 提示词保留实际使用时的英文原文。

Create a GIF. Create a single animation sprite sheet for a seamless looping GIF. A cheerful round orange fox mascot wearing a small teal scarf stands facing the viewer. The sequence: looks up, notices the viewer, gives a big wave, hops once with joy, lands, then settles back into its starting pose. Soft 3D toy-like style, clean studio lighting, plain warm cream background. Exactly 6 sequential frames in a precise 3-column by 2-row grid, read left to right, then top to bottom. Every panel has identical dimensions. Fixed camera, same character, same proportions, same background. Clear incremental motion; the sixth frame transitions smoothly back to the first. No text, logos, panel labels, borders, or gutters.

Create a GIF. Create a single animation sprite sheet for a seamless looping GIF. A premium unbranded matte-black insulated bottle stands on a wet dark stone surface. The sequence: a water droplet falls from above, hits the surface beside the bottle, ripples spread outward, small droplets rise and catch the light, then the surface calms again. The bottle stays perfectly still and identical in every frame. Luxury product photography, dramatic rim lighting, dark teal background. Exactly 6 sequential frames in a precise 3-column by 2-row grid, read left to right, then top to bottom. Every panel has identical dimensions. Fixed camera. Clear incremental motion; the sixth frame transitions smoothly back to the first. No text, logos, panel labels, borders, or gutters.

Create a GIF. Create a single animation sprite sheet for a seamless looping GIF. A tiny sleepy penguin in an oversized knitted sweater sits at a small table. The sequence: dozes off with heavy eyelids, a steaming cup of coffee slides into view, one eye opens, it takes a sip, its eyes go wide with sparkles, then it slowly starts dozing again. Expressive cute 3D animation style, soft morning light, cozy pastel kitchen background. Exactly 6 sequential frames in a precise 3-column by 2-row grid, read left to right, then top to bottom. Every panel has identical dimensions. Fixed camera, same character, table, and background. Clear incremental facial motion; the sixth frame transitions smoothly back to the first. No text, logos, speech bubbles, panel labels, borders, or gutters.

起始草图(同样由 AI 生成)

动画效果
Create a GIF. Using the attached hand-drawn sketch as the character design, create a single animation sprite sheet for a seamless looping GIF. Keep the character's shape, proportions, and hand-drawn pen line quality, and keep the lined paper background. The sequence: the character blinks, looks left, looks right, does a small happy jump, lands, then returns to its starting pose. Exactly 6 sequential frames in a precise 3-column by 2-row grid, read left to right, then top to bottom. Every panel has identical dimensions. Fixed camera, same character, same background. Clear incremental motion; the sixth frame transitions smoothly back to the first. No added text, panel labels, borders, or gutters.

Create a GIF. Create a single animation sprite sheet for a seamless looping walk cycle. A small pixel-art knight with a blue cape and a round shield walks to the right, seen from the side. Crisp 32-bit pixel art with a limited color palette and a plain light gray background. Exactly 8 sequential frames in a precise 4-column by 2-row grid, read left to right, then top to bottom. Every panel has identical dimensions. The character stays centered in each panel at the same scale, with feet on the same baseline. Clear leg, arm, and cape motion; the eighth frame transitions smoothly back to the first. No text, panel labels, borders, or gutters.
这三张来自同一段 ChatGPT 对话。人物的脸、发型、眼镜和外套都延续了下来,但细看最后一张,疤痕和狐狸徽章的位置变了。

A portrait of a young woman explorer with short copper hair, round brass glasses, a small scar on her left eyebrow, and a mustard-yellow utility jacket with a green fox patch on the chest. Semi-realistic digital illustration, neutral gray background, waist-up view.

Keep the same character exactly: same face, hair, glasses, scar, jacket, and fox patch. Place her in a rainy neon-lit night market, holding a paper umbrella. Change only the setting, pose, and lighting.

Keep the same character exactly: same face, hair, glasses, scar, jacket, and fox patch. Place her on a snowy mountain ridge at sunrise, looking through binoculars. Change only the setting, pose, and lighting.
每一行左边来自 OpenAI 公告,右边是本站生成器目前的实际情况,做不到的也会写明。
GPT Image 2.6
OpenAI 称生成速度比上一版本提升 50%。
AIGPTImage 现状
本站默认使用 2.6 Flare。单次最多可生成 4 张,具体取决于你的套餐。
GPT Image 2.6
新增 GIF 输出,适合吉祥物、表情包和分镜等用途。
AIGPTImage 现状
把输出格式选成 GIF,就能用文字生成 2 到 16 帧、约 512 像素的循环动图。需要更长的动态画面时,可以用本站的 Seedance 视频生成器让静态图动起来。
GPT Image 2.6
多次编辑时,人物、产品和品牌细节会更稳定。
AIGPTImage 现状
可以上传最多 16 张参考图,并写明哪些内容不能变。2.6 Sunburst 更注重多轮编辑的一致性,但每一轮仍建议核对小细节。
GPT Image 2.6
原本只在 ChatGPT 中提供的图片标注,现在也进入了 API。
AIGPTImage 现状
本站没有标注工具。请用文字说明位置,例如“只改左边瓶子上的标签”。
GPT Image 2.6
设计稿可以导出为 SVG。
AIGPTImage 现状
不支持矢量导出。下载的是位图,可选 1K、2K 或 4K,具体取决于你的套餐。
GPT Image 2.6
API 新增 complexity 参数,用来决定模型为请求做多少规划,与画质设置相互独立。
AIGPTImage 现状
本站没有对应设置,可以调整的是宽高比、分辨率和输出数量。
上方生成器里两款 2.6 模型都可以选,可以先看手头的任务。
| 你的需求 | 建议先试 | 原因 |
|---|---|---|
| 快速出一批草稿 | Flare | OpenAI 表示,低画质下最快约 6 秒出图。 |
| 海报和产品视觉 | Flare | OpenAI 特别提到它改进了这类专业视觉的版式。 |
| 细节丰富、要求精确的画面 | Sunburst | OpenAI 把它定位为画质更高的模型。 |
| 反复编辑同一主体 | Sunburst | 中高画质下,它在多次编辑之间的一致性更强。 |
OpenAI 还建议:多数编辑用低 complexity,元素多或限制多的场景用中或高。ChatGPT 和 Codex 的所有订阅档位都能使用 Images 2.6。
在本站,Flare 和 Sunburst 消耗的积分相同。2.5 版本仍然可选,积分更少。
这几条是为本页生成器写的。每条都对构图或细节保持提出了明确要求,出图后容易看出问题。
竖版海报,主题是社区小型爵士之夜。上方三分之一放大号衬线标题“蓝调时分”。标题下方,一把立式贝斯靠在砖墙边,头顶一盏暖色路灯。底部居中写“每周五晚八点”。配色为深靛蓝和琥珀色,留出宽边距,不要其他文字。
使用这条提示词商品目录照片:一只带斑点的白色陶瓷咖啡滤杯放在同款马克杯上,略带四分之三侧角。浅鼠尾草绿背景,左侧一盏柔光,右侧有淡淡的投影。釉面斑点要清晰,比例要真实。不要品牌标识、道具和文字。
使用这条提示词一张图排成四格分镜,上下各两格,格间留细白缝。雨天城市路边的水沟里有一只折纸船:先出发,然后在小漩涡里打转,再从落叶下穿过,最后驶入倒映霓虹招牌的大水洼。每格都是同一只纸船、同样的水彩风格和相同的机位高度。不要说明文字和数字。
使用这条提示词每条提示词都生成静态图。分镜是一张包含四格的图片,不是动画。
可以。生成器默认使用 GPT Image 2.6 Flare,也可以选 2.6 Sunburst,文生图和图生图都支持。
可以。2.5 Flare 和 2.5 Sunburst 仍在模型列表里,积分更少:1K 6 积分,2K 10 积分,4K 16 积分。
GIF 可以:选择 GPT Image 2.6,把输出格式设为 GIF。GIF 约 512 像素(1:1 为 512×512),仅支持文生图,8 帧以内 10 积分,9 到 16 帧 20 积分。不支持导出 SVG,下载的是位图文件。
草稿、海报和产品图先用 Flare;细节要求高或反复编辑同一主体时选 Sunburst。在本站两者价格相同。
GPT Image 2.6 每张图在 1K 下消耗 10 积分,2K 消耗 20 积分,4K 消耗 34 积分。图生图包含 1 张参考图,每多一张加 2 积分。1K 透明背景不额外收费。较高分辨率可能取决于你的套餐。
不是。AIGPTImage 是独立产品,并非由 OpenAI 运营或背书。