GPT-Image-2 is OpenAI’s image generation model, available on Labnana at 1K, 2K, and 4K. It is characterised by strong instruction adherence and clean illustrative rendering.

How do I use these GPT-Image-2 prompts?

Select Generate on any prompt card to pre-fill the generator with that model and prompt, or copy the prompt text for use elsewhere.

Is GPT-Image-2 free on Labnana?

Free GPT-Image-2 generations are issued through platform campaigns, and the model dropdown shows how many remain on your account. Beyond that it bills against universal credits, which a free account accrues at 5 per daily check-in. 4K requires a paid plan.

How should a GPT-Image-2 prompt be structured?

Specify subject, composition, style, and lighting as separate requirements. The model follows structured instructions closely, so detail improves adherence. Comparing any card’s text against its image is the fastest way to calibrate.

GPT-Image-2 or Nano Banana?

Use GPT-Image-2 when the output must match a written spec and the style is illustrative. Use Nano Banana for in-image typography, character consistency, and photorealistic detail. Both run on Labnana, so the practical answer is to run one prompt on each.

Can GPT-Image-2 render text inside images?

Short, clearly quoted text is handled well. For dense or heavily stylised typography, Nano Banana Pro is the stronger option.

Which resolutions does GPT-Image-2 support?

1K, 2K, and 4K. 1K and 2K are available on the free plan; 4K requires a paid plan.

Are these prompts free to use?

Yes — the prompts are community-published and free to copy and adapt.

Can the output be used commercially?

Labnana applies no watermark to your generations. Before using images in paid work, review the model providers’ usage policies.

1 imageGPT-Image-2

Home/Models/GPT-Image-2

GPT-Image-2 · OpenAI

GPT-Image-2 Prompts

OpenAI’s image model on Labnana, indexed with the prompts that produced each result.

See prompt examples

Literal instruction following
Clean illustrative rendering
Reference-guided generation
4K output

Introduce Safranbolu, half text and half image, blending together

Theme:
“Stay in a UNESCO World He — Real GPT-Image-2 generations — open one for the full prompt

Cappadocia Cappadocia

“The world's only fantasy landscape suitable for viewing from a hot air ball — Real GPT-Image-2 generations — open one for the full prompt

GPT-Image-2 is OpenAI’s image generation model, available on Labnana at 1K, 2K, and 4K. Its distinguishing property is instruction adherence: a brief that specifies subject, composition, style, and lighting is rendered close to what was written, which makes it the dependable choice when the output has to match a spec rather than surprise you.

Every prompt below is taken from a published generation, paired with the image it produced. Load one into the generator at the top of the page and it runs on GPT-Image-2 with the original parameters intact.

Using GPT-Image-2 on Labnana

Two channels in the generator

GPT-Image-2 appears under OpenAI in the model dropdown, alongside a Stable channel to fall back on when run-to-run consistency matters more than variation.

Write a brief, not a subject

The model rewards structure. Specify subject, composition, style, and lighting as separate requirements and each one is carried into the output rather than averaged away.

Resolution and aspect ratio

1K, 2K, and 4K across eight ratios from 1:1 to 21:9. 4K requires a paid plan; 1K and 2K are open on the free plan.

Reference images

Attach up to 5 references per generation to carry a style, subject, or layout forward instead of re-describing it in every prompt.

Free quota on GPT-Image-2

GPT-Image-2 free generations are issued through platform campaigns; the model dropdown shows the remaining count on your account in real time.

Campaign quota: Granted per campaign — remaining count shown in the generator
Daily check-in: 5 universal credits, usable on any model
Free plan access: 1K and 2K — free quota first, universal credits after
Resolution limit: 4K requires a paid plan; free quota never applies to 4K
Watermarking: None, on free or paid plans

For a standing free allowance that does not depend on a campaign, Nano Banana Pro is the model to use: 2 generations on sign-up and +1 from each daily check-in.

View plans and credits

Trending GPT-Image-2 Prompts

Ranked by current engagement across published GPT-Image-2 generations.

GPT-Image-2

Introduce Safranbolu, half text and half image, blending together Theme: “Stay in a UNESCO World Heritage Site”View full prompt

GPT-Image-2

Cappadocia Cappadocia “The world's only fantasy landscape suitable for viewing from a hot air balloon” Keywords: Fairy Chimneys Love Valley Pigeon Valley Göreme Open Air Museum View full prompt

GPT-Image-2

{ "画面基础设定": { "画风": "经典魔法少女赛璐璐手绘风格，平涂上色，干净利落手绘线条，无厚涂、无写实3D质感，4K高清，16:9画幅，光影柔和统一无明暗突变，人物无崩坏、无透视穿模", "场景背景": "华丽欧式王座大厅，画面深处摆放雕花复古王座，整体氛围为守护向，核心叙事：蓝裙洛丽塔少女保护婚纱少女，暗黑水手服少女对峙二人", "构图氛围": "中景人物互动特写，景深虚化背景突出三位少女主体，色调冷暖对比，暗黑水手服少女冷暗色调，洛丽塔与女王柔和暖调" }, "角色详细动作与形象": [ { "角色编号": "角色1（图一蓝发洛丽塔少女）", "动作姿态": "站在画面中间位置，处于角色2与角色3中间形成阻隔；双手握持概念剪刀「织命」横举在身前，剪刀刀刃朝向角色2，身体微微侧转面向角色2，身体向身后的角色3倾斜，做出保护姿态，眼神警惕直视前方水手服少女" }, { "角色编号": "角色2（图二暗黑水手服少女）",右手佩戴黑色毛绒猫爪手套，左手为人类纤细素手", "动作姿态": "站在画面左侧，身体正对中间的角色1；抬起佩戴猫爪手套的右手，指尖直指角色1方向，左手自然垂落身侧，眼神冷冽紧盯对面两名少女" }, { "角色编号": "角色3（图三婚纱女王）", "动作姿态": "站在角色1正后方，从身后伸出双臂环抱住角色1；双手在角色1脖颈前方交叉交织紧扣，下巴轻轻搭在角色1的肩膀上，侧脸贴靠角色1后背，姿态依赖柔和；人物身后清晰露出欧式雕花王座轮廓" } ], "道具说明": { "道具名称": "织命剪刀", "外观": "高碳钢镜面刀身，深蓝色树脂握柄镶嵌银色卷草纹样，长短双指环结构，由角色1双手紧握横挡身前作为防御道具" }, "画面核心叙事": "蓝洛丽塔少女手持剪刀隔开对峙的水手服少女，用身体护住身后被女王环抱的自己，王座衬托女王的特殊身份，整体画面传递少女挺身守护同伴的核心故事感" } 请规避繁复的细微纹理、过度的细节刻画、颗粒噪点、粗糙质感、斜向网格纹路、瓷砖式重复纹样、泥污类叠加效果，同时去掉多余的闪光特效。弱化复杂纹理质感，优先保证画面表面顺滑均匀、轮廓线条粗实清晰、主体轮廓辨识度高。相较于堆砌细节，更注重造型利落、留白恰当、构图规整，打造干净的画面观感。View full prompt

GPT-Image-2

Overall layout is a concept design sheet — Magical Girl · Iris. The concept design background is light grid paper. Color scheme follows Figure 1 and Figure 2. Three-view character turnaround sheet: front view, left side view, back view. Show clothing details. Use Chinese. The weapon is a cross-shaped long two-handed staff. The back view includes a large back bow that matches the front view; both the white inner hem and blue outer hem are slit at the back, and the blue outer hem is shorter than the white inner hem, matching the length shown in the front view. The nun's veil has 1 layers, and its length is 2/3 of the hair. Hair length reaches the waist.View full prompt

GPT-Image-2

Replace the characters from image 2 and image 3 completely into the scene of image 1, strictly maintaining the body posture, facial expression, gaze angle, and all subtle movements of the characters in image 1 without any deviation. The clothing should remain consistent with image 2 and image 3 respectively, with image 2 on the left and image 3 on the right. Ensure that after replacement, the characters fully integrate with the scene's lighting, perspective, and compositional atmosphere, avoiding any sense of disjointed layering or perspective errors. Focus on optimizing the proportions and symmetry of the characters' facial features, especially the perspective of the eyes and the position of the pupil highlights, which must precisely correspond to the expressions in the original images. Also, carefully handle the structure of the fingers and joint details to prevent common issues such as extra fingers, fused fingers, or structural distortion. The final result should be a highly cohesive anime-style image with clear and smooth character lines, natural color transitions, no obvious signs of artificial generation, and an overall quality resembling hand-drawn artwork by an original artist. Ensure that the characters' actions and expressions remain unchanged. The actions should fully reference image one. proportions should maintain those of image 1, with a resolution of 4K.View full prompt

GPT-Image-2

Weapons and clothing are more ornate, beautiful, and futuristic, with a scorpion tail added. The overall clothing style remains unchanged, with venomous stinger elements. The overall color scheme is black and green.View full prompt

GPT-Image-2

[Overall Style] vertical poster, semi-painted thick coating combined with letterpress printing texture, full color. The overall image uses sky blue, cream white, and orange red as the main tones, presenting a 1970 era printed poster style. The theme is a soda poster. The entire image is covered with a fine layer of grain noise, evoking a nostalgic atmosphere of the era. [Composition Design] In the center of the image is the figure of the 1 character from the legs upward, occupying the visual center. The upper part of the image features a sky blue diagonal strip, with white retro advertising font reading: “ Arctic Ocean Soda ”. In the lower right corner, in white advertising art font with a sky blue outline, it reads “ Arctic Ocean™ ” in small text. [Character Details] Positioned in the center of the image, leaning the body slightly and lifting one leg in an air-sitting pose, one hand holding a Arctic Ocean glass bottle filled with yellow soda (as shown in figure 2) raised to the cheek, the other hand giving a thumbs up and extending toward the camera. The character's eyes are bright, playfully closing one eye, with a sunny and confident smile. The gaze looks toward the camera. [Color and Texture] Cream white is the main background color. The entire image is covered with grain noise, with white and blue color blocks having a rough printed texture, fading the vividness, presenting a nostalgic atmosphere of the era.View full prompt

GPT-Image-2

Extract character, anime CG, with a slight hand-drawn style. Capture the moment the character looks at the camera from the side. Character height 148 cm The pendant accessory on the character's chest is three-dimensional and spherical. Vividly showcase the character's temperament. No text appears in the image.View full prompt

GPT-Image-2

Draw a close-up portrait of the character from Figure 1 as the main reference. The image is composed of screens from various electronic devices of different eras, pieced together like a fragmented portrait collage made of screens. The character is based on Figure 1, retaining her most representative facial features, hairstyle, temperament, and white veil covering her eyes. She sits upright, smiling at the camera, with an overall atmosphere that is mysterious, quiet, with a slight technological feel and a sense of being watched. The main subject of the image is a complete close-up of the character's face, but this face is not directly drawn on a single canvas; instead, it is split and displayed across the screens of multiple electronic devices. The devices include feature phones, flip phones, early smartphones, modern smartphones, MP4 players, pagers, handheld game consoles, small digital players, and so on. Due to differences in era, screen material, and display capabilities, the image quality varies across devices: feature phones and pagers have lower screen resolution, possibly showing pixelation, noise, color shifts, scan lines, low saturation, or incomplete display; flip phones and MP4 screens are slightly blurry; modern smartphone screens are clearer with more accurate colors. Prohibited: irregular screen shapes, non-standard screens, triangular screens, circular phone screens, floating screens, transparent screens, sci-fi fictional screens, screens without bezels, devices that do not match reality, excessive deformation of devices, curved phones, curved screens, garbled text, excessive text, unreadable text, face completely obscured by text, screens that cannot be pieced together at all, collage that does not form a human face, all screens being the same type of device, the outer area filled with devices, no blank space left.View full prompt

GPT-Image-2

以第一张三视图和第二张证实图为角色基础，第三四为同角色同人图，原型为在中国的哔哩哔哩视频网站的虚拟偶像“宣小纸不怕火”，个人主页为https://space.bilibili.com/3461582034045213。自动生成一张收藏版史诗叙事海报：巨大优雅的人物侧脸剪影作为外轮廓，剪影内部自动生长出最契合该主题的完整世界观、标志性场景、角色关系、象征符号、关键建筑、生物、道具与氛围。整体不是普通拼贴，而是高级的剪影轮廓填充式叙事合成，带有双重曝光式联想，但更偏电影海报与梦幻水彩插画融合风格；柔和空气透视，轻雾化过渡，纸张颗粒，边缘飞白与刷痕，大面积留白，版式克制高级，安静、宏大、神圣、怀旧、诗意、传说感强。风格、色彩、场景、材质全部根据主题自动适配，所有元素必须强绑定主题，一眼识别，不要杂乱，不要硬拼贴，不要模板化背景，不要廉价奇幻素材。画面中需自然加入专属签名“【宣小纸不怕火】”，作为海报设计的一部分，位置低调但清晰，可放在左下角、右下角或标题附近，风格需与整体版式统一，像收藏版海报的作者落款或设计签章；签名字体精致、克制、高级，不可过大，不可破坏主体构图，不可显得突兀廉价。分辨率为2K View full prompt

How it works

Start from a verified result

Each card pairs a published GPT-Image-2 image with the full prompt that produced it.

Load it into the generator

Generate pre-fills the prompt and switches the model to GPT-Image-2. The raw text can also be copied for use elsewhere.

Iterate one variable at a time

Run it unmodified first, then change a single requirement so the effect of each edit stays attributable.

About GPT-Image-2 and Prompts

This library indexes GPT-Image-2 generations published on Labnana together with their prompts. Nothing here is written speculatively: if an image appears on this page, that exact text produced it.

GPT-Image-2 is OpenAI’s image generation model. On Labnana it sits alongside Google’s Nano Banana family and ByteDance’s Seedream 5.0 Pro, and the same prompt can be re-run on any of them for comparison.

Generate with GPT-Image-2

Other models on Labnana

All models

GoogleNano Banana PromptsIn-image text, character consistency, photoreal detail

ByteDanceSeedream 5.0 Pro PromptsRegion-level edits, local corrections, Chinese instructions

Frequently asked questions

Start with GPT-Image-2

Load any prompt on this page into the generator, or write your own.

View pricing

Labnana

AI image and video generation, with models from Google, OpenAI, ByteDance and Alibaba.

Features

AI Image Generator
AI Video Generator
Explore
Pricing
API Docs
All Models

Image Models

Nano Banana
GPT-Image-2
Seedream 5.0 Pro

Video Tools

Image to Video
Text to Video
Reference to Video

Video Models

Seedance 2.0
HappyHorse 1.1

Using GPT-Image-2 on Labnana

Two channels in the generator

Write a brief, not a subject

Resolution and aspect ratio

Reference images

Free quota on GPT-Image-2

Trending GPT-Image-2 Prompts

How it works

Start from a verified result

Load it into the generator

Iterate one variable at a time

About GPT-Image-2 and Prompts

Other models on Labnana

Frequently asked questions

What is GPT-Image-2?

How do I use these GPT-Image-2 prompts?

Is GPT-Image-2 free on Labnana?

How should a GPT-Image-2 prompt be structured?

GPT-Image-2 or Nano Banana?

Can GPT-Image-2 render text inside images?

Which resolutions does GPT-Image-2 support?

Are these prompts free to use?

Can the output be used commercially?

Start with GPT-Image-2