比Nano Banana还夯的生图模型泄露，截图不再是证据了 | 附提示词

marsbitPublicado em 2026-04-19Última atualização em 2026-04-19

Resumo

近日，LM Arena评测平台短暂出现了三个匿名图像模型，据推测为OpenAI尚未正式发布的GPT Image 2。该模型在文字渲染、指令跟随、真实感和世界知识等方面表现突出，尤其在多语言文字（包括中文）的生成上准确率大幅提升，能生成高度逼真的证件、界面和产品标签图像。 这一进步使得截图作为证据的可信度受到挑战，但也为设计、产品原型制作及内容配图等场景带来便利。与此同时，GPT Image 2在多项测试中表现优于Google的Nano Banana Pro及Midjourney等模型。 目前该模型仍处于A/B测试阶段，预计将在5月随DALL-E系列服务退役后正式发布。用户可尝试在LM Arena平台体验，并参考社区推荐提示词生成高质量图像。

你对文生图的印象还停留在 Nano Banana 吗?

可是孩子,时代又变了。

@johnAGI168 https://x.com/johnAGI168/status/2044781168151724067

@0115hippo https://x.com/0115hippo/status/2044722124611539160

4 月初,LM Arena 评测平台上出现了三个匿名图像模型,代号分别是 maskingtape-alpha、packingtape-alpha、gaffertape-alpha。几小时后它们消失了。

OpenAI 官方还没有正式宣布这个模型,但根据 API 返回的元数据和用户侧的测试记录,它已经有了一个被广泛接受的名字:GPT Image 2。

截图不能再当证据了

过去几年,AI 生图模型最明显的短板之一就是图片里的文字。DALL-E 3 时代,你让它在图里写「Hello」,出来的可能是「Hellp」甚至「Hl10」,字母像喝醉了一样东倒西歪。GPT Image 1 好了很多,能处理简单的英文标签。到 GPT Image 1.5 ,其对英文文字的渲染准确率已经接近 95%,但在中文、日文、韩文等非拉丁字母体系上仍有明显缺陷。

而 GPT Image 2 的泄露样图改变了这个印象。

@MrLarus https://x.com/MrLarus/status/2044824800909054181

@akokoi1 https://x.com/akokoi1/status/2044789531615056175

图片里的文字,该是什么就是什么。中文清晰,字形准确,笔画完整。有人测试生成一张身份证样式的图片,姓名、地址、证件号码全部正确渲染,排版规整,初看像是真实文件的照片。

这是个好消息。文字渲染的进步,意味着生成信息图、海报、产品包装、排版复杂的图表,都变得更可靠了。

但硬币总有另一面。一个能生成以假乱真的证件样式图、精确渲染 UI 截图的模型,自然也让「截图可以作为证据」这件事变得越来越可疑。

对比来看,这也是 GPT Image 系列和其他模型的核心差异所在。Midjourney 至今在文字渲染上毫无建树,Stable Diffusion 系列也是老问题。根据泄露的 Arena 测试结果,GPT Image 2 在文字渲染、指令跟随、照片真实感和世界知识四个维度上均超过 Midjourney,后者的优势主要保留在艺术风格和美学控制上。

它真的知道这个世界长什么样吗

有测试者让模型生成一个假想的 GPT-8 产品定价页面,结果出来的图,排版确实是 OpenAI 官网的风格,按钮位置和字体选用像是从真实界面截取的,价格表格的层级逻辑也是对的。

GPT Image 2 能生成与真实软件界面极为相似的图像,包括浏览器窗口、移动端应用界面、数据可视化图表,保真度是上一代产品无法比拟的。

@johnAGI168 https://x.com/johnAGI168/status/2044781168151724067

@levelsio https://x.com/levelsio/status/2040333489476681758

这将带来一些很有意思的实际用途。设计师在做产品原型的时候,不需要先打开 Figma 画一堆框架,直接用文字描述想要的界面,出来的就是一张可以用来和团队讨论的参考图。做投资人 Deck 时,不需要等工程师写代码就能展示一个「产品截图」。写文档的时候,用来配图的示例界面可以直接生成,不用对着空白页面想截图从哪里找。

@marmaduke091 https://x.com/marmaduke091/status/2040338311873515597

生图这件事,已经不只是「生图」了

OpenAI 已经宣布 DALL-E 2 和 DALL-E 3 将于 2026 年 5 月 12 日正式停止服务。Azure OpenAI 的 DALL-E 3 已经在 2 月提前退役了。

DALL-E 是很多人第一次接触 AI 生图的地方,从那些模糊的早期作品到今天,才短短几年。

与此同时,2026 年初刚刚凭借 Nano Banana Pro 确立行业地位的 Google,或许将感受到压力。早期测试报告显示,GPT Image 2 在真实感、文字渲染和世界知识三个维度上同时超越了 Nano Banana Pro,这种三连胜并不常见。

对于创作者来说,感受是复杂的。插图师、平面设计师、摄影师,已经不是第一次面对这个话题了。自 GPT Image 1 发布以来,自由职业平面设计职位数量下降了约 18%。AI 在某些场景下确实取代了「我要雇一个人做这件事」的决策,但它也在创造新的工作方式,让一个人能做的事变多了。

生图模型的进化速度,已经不再给人留出太多适应时间了。GPT Image 1 从上线到 1.5,不过几个月。1.5 到 2,大概也就半年。每一代都在解决上一代的核心短板,同时打开新的可能性。

GPT Image 2 现在还处于 A/B 测试阶段,部分 ChatGPT 用户已经随机获得了访问权限。正式发布的时间窗口,普遍预测就在 5 月 DALL-E 退役前后。想要提前体验的话,目前可以在 LM Arena 评测平台碰碰运气。

Test Address: https://arena.ai

根据社区反馈和该模型的已知优势,以下提示模板可以最大限度地提高你的成功几率:

UI/截图提示:一张照片级逼真的手机银行应用截图,清晰显示交易记录,其中日期、金额和商户名称清晰可辨。iPhone 16 屏幕,自然手持手机,咖啡店背景。

产品标签提示:一张照片级精酿啤酒瓶产品照片,标签细节清晰,显示酒厂名称「Oakridge Brewing Co.」,酒精度 6.8%,山脉标志及配料表。棚内布光,白色背景。

标识提示:一张东京夜间巷道的街景照片,可见多处日英双语霓虹灯招牌,包括写有「Ichiban Ramen — Est. 1987」的拉面店招牌、卡拉 OK 酒吧招牌以及各种发光的广告牌。雨后湿滑的人行道上映着灯光。

界面/世界知识提示:一张照片级真实的 YouTube 视频截图,展示了一段名为「如何在 2026 年组装电脑」的视频,该视频拥有 230 万次观看,配有逼真的评论区、侧边栏推荐视频以及频道信息。桌面浏览器视图。

宽屏触发提示:这是一张电影般的宽银幕照片,拍摄了宜家门店黄昏时分的外观,展示了发光的宜家招牌、停车场里有逼真的汽车,以及进进出出的购物者。黄金时刻灯光,格式 16:9。

未标注图片来源及参考:https://miraflow.ai/blog/how-to-use-duct-tape-ai-model-arena-gpt-image-2-guide

本文来自微信公众号“APPSO”,作者:发现明日产品的

Criptomoedas em alta

ctr
CitreaCTR
wstusdt
wrapped stUSDTWSTUSDT
apr
aPrioriAPR
ctx
Cryptex FinanceCTX
audio
AudiusAUDIO
mantra
MantraMANTRA
xvs
VenusXVS
waxl
AxelarWAXL
bill
Billions NetworkBILL
pyth
PYTH (Pyth)PYTH
rune
THORChainRUNE
velodrome
Velodrome FinanceVELODROME
brev
BrevisBREV
zrx
ZRX（0X）ZRX
cake
PancakeSwapCAKE

Perguntas relacionadas

QGPT Image 2相比前代模型在文字渲染方面有哪些显著提升？

AGPT Image 2在文字渲染方面实现了显著突破，能够准确生成中文、英文等多种语言的文字，字形和笔画完整清晰。例如，它可以生成排版规整的身份证样式图片，甚至能正确渲染复杂的UI界面文字，解决了早期模型（如DALL-E 3）文字错乱的问题。

QGPT Image 2的泄露对‘截图作为证据’的可靠性有何影响？

AGPT Image 2能够生成高度逼真的证件、UI界面等图像，且文字渲染极其准确，这使得伪造的截图难以被肉眼识别。因此，传统上依赖截图作为证据的可靠性大幅降低，因为AI生成的虚假截图可能以假乱真。

QGPT Image 2在哪些实际应用场景中具有优势？

AGPT Image 2在生成信息图、产品原型设计、投资演示、文档配图等场景中优势明显。例如，设计师可以直接用文字描述生成界面参考图，无需手动绘制；投资者能快速生成产品截图用于Deck展示，提高效率。

QGPT Image 2与Google的Nano Banana Pro相比如何？

A根据早期测试，GPT Image 2在真实感、文字渲染和世界知识三个关键维度上均超越了Google的Nano Banana Pro，实现了全面领先。而Nano Banana Pro仅在艺术风格控制上可能保留部分优势。

Q如何提前体验GPT Image 2模型？

A目前GPT Image 2处于A/B测试阶段，部分ChatGPT用户可能随机获得访问权限。用户也可以尝试通过LM Arena评测平台（https://arena.ai）碰运气，但需注意该模型尚未正式发布，OpenAI计划在2026年5月DALL-E退役前后推出。

Leituras Relacionadas

South Korea Moves to Regulate Cross-Border Crypto Transfers Under New Framework

South Korea is expanding its regulatory framework for cross-border virtual asset transfers, set to launch in December, by including fintech companies. Under the new rules, firms conducting such transfers must register with the Ministry of Economy and Finance and report transactions through the foreign exchange system. The move aims to bring previously unsupervised crypto transfers under formal oversight to address money laundering and crime risks. While initial expectations were that major crypto exchanges like Upbit would dominate the new licensing system, regulators now plan to extend eligibility to non-traditional entities, including fintech firms, if they can efficiently perform transfers. Authorities are finalizing implementation rules with industry stakeholders ahead of the December launch. This framework is part of broader efforts to strengthen digital asset oversight in South Korea, which includes developing rules for tokenized securities, potentially subjecting them to securities taxes.

TheNewsCryptoHá 1h

South Korea Moves to Regulate Cross-Border Crypto Transfers Under New Framework

TheNewsCryptoHá 1h

Matt Damon To Speak At Ripple Swell As Water.org’s RLUSD Push Draws Attention

Matt Damon is scheduled to deliver a keynote at Ripple Swell 2026, representing the non-profit Water.org which he co-founded. His appearance highlights a partnership between Water.org’s "Get Blue" campaign and Ripple, the campaign’s exclusive digital asset and payments partner. Ripple will utilize its payment network and RLUSD stablecoin to facilitate faster, lower-cost cross-border transactions for the campaign’s microfinance partners. This collaboration provides RLUSD with a mainstream, humanitarian use case focused on operational efficiency in aid funding, moving beyond typical stablecoin narratives centered on trading and finance. For Ripple, the partnership offers a reputational boost by framing its technology as practical payment infrastructure for social impact, though the actual adoption and scale of such use cases remain to be seen.

bitcoinistHá 1h

Matt Damon To Speak At Ripple Swell As Water.org’s RLUSD Push Draws Attention

bitcoinistHá 1h

Microsoft Identifies New Crypto Malware Targeting Wallet Addresses and Private Keys

In February 2026, Microsoft identified a new crypto clipper malware, dubbed Trojan/CryptoBandits.A, targeting Windows systems. The malware spreads via malicious shortcut files on USB drives and operates without a traditional installer or control servers by leveraging Windows Script Host and ActiveX to deploy a Tor proxy. Once active, it runs two modules: one for spreading and another for stealing information. The malware continuously monitors the clipboard for 12 or 24-word recovery phrases, Bitcoin/Ethereum private keys, and wallet addresses. When a user copies a wallet address, the malware silently swaps it with one controlled by attackers to divert funds. It also captures screenshots to gather information on wallet balances and user activity, sending data through Tor connections. Additional capabilities include remote code execution and persistence via scheduled tasks. Microsoft advises disabling auto-run features, restricting script interpreters and executable shortcuts from USB drives, and monitoring for suspicious activities like JavaScript execution, localhost:9050 proxy use, PowerShell screenshot capture, and clipboard monitoring.

TheNewsCryptoHá 1h

Microsoft Identifies New Crypto Malware Targeting Wallet Addresses and Private Keys

TheNewsCryptoHá 1h

No Sales Team, $20 Million in Revenue: How Did AI Employee Viktor Win Over 30,000 Companies?

The AI employee Viktor, developed by a team with DeepMind background, has achieved $20 million in annual revenue without a traditional sales team, serving over 30,000 companies. Its core innovation lies in positioning itself as a "Tier 3 AI Coworker" capable of "end-to-end execution and delivery of results," moving beyond the "draft and wait for human completion" model of typical AI assistants. Users can simply mention Viktor in Slack or Microsoft Teams using natural language commands, and it autonomously performs tasks like pulling sales data from a CRM, generating reports, or even cross-tool operations like creating board meeting PPTs by aggregating data from six different sources. Key to its growth is a pure Product-Led Growth (PLG) model, eliminating complex implementation cycles and per-seat licensing. Instead, it charges based on task credits or consumption, lowering the trial barrier with a $100 free credit offer and no credit card required. This enabled viral, bottom-up adoption within organizations. Viktor's interaction paradigm removes the barrier of prompt engineering, allowing non-technical employees to delegate complex workflows seamlessly. It also features proactive, automated task execution (e.g., overnight bookkeeping, scheduled reports) based on triggers, effectively embedding AI as an automated "process layer" within business operations. However, its expansion into Microsoft Teams—a platform with 320 million users—highlights challenges. Large enterprises require stringent IT compliance, security reviews (e.g., SOC 2), and governance, potentially hindering the frictionless, user-driven adoption that succeeded in Slack. Additionally, the "black box" nature of its autonomous decision-making raises concerns about operational risks, data integrity, and the need for robust audit logs and permission controls. Balancing efficiency gains with security and trust remains a critical hurdle for Viktor and similar AI agents aiming to become core enterprise infrastructure.

marsbitHá 2h

No Sales Team, $20 Million in Revenue: How Did AI Employee Viktor Win Over 30,000 Companies?

marsbitHá 2h

Interview with CoreWeave Co-founders: AI Demand Seems to 'Intensify' Every Day

An Interview with CoreWeave Executives: AI Demand Seems to 'Intensify' Every Day In an interview, CoreWeave executives highlight a structural shift in AI infrastructure demand. While GPU availability remains crucial, the primary bottlenecks are evolving to include powered data center shells, skilled labor (like electricians), and complex supply chain execution. They note that AI demand, particularly for agentic AI and reasoning models, continues to intensify daily, accelerating since Q1 2024. This demand is driving a need for more balanced infrastructure. CoreWeave is redesigning data centers to allocate more space for storage and CPUs alongside GPUs, with significant interest in Nvidia's upcoming Vera CPUs. The company, serving top AI labs and hyperscalers, emphasizes its client-driven model, building precisely to customer specifications. CoreWeave attributes its competitive edge to proven execution, performance, and a mature platform for AI deployment. Pricing is structured to pass component cost increases (e.g., for HBM memory) to customers, protecting margins. Looking ahead, they anticipate Vera Rubin platform deployments to begin meaningfully in late 2025, with a major ramp throughout 2027, mirroring the Blackwell (GB) series rollout pattern. The competition is shifting from merely acquiring chips to holistic engineering and delivery capability.

marsbitHá 2h

Interview with CoreWeave Co-founders: AI Demand Seems to 'Intensify' Every Day

marsbitHá 2h

Trading

Spot
Futuros

Artigos em Destaque

Como comprar BANANA

Bem-vindo à HTX.com!Tornámos a compra de Banana Gun (BANANA) simples e conveniente.Segue o nosso guia passo a passo para iniciar a tua jornada no mundo das criptos.Passo 1: cria a tua conta HTXUtiliza o teu e-mail ou número de telefone para te inscreveres numa conta gratuita na HTX.Desfruta de um processo de inscrição sem complicações e desbloqueia todas as funcionalidades.Obter a minha contaPasso 2: vai para Comprar Cripto e escolhe o teu método de pagamentoCartão de crédito/débito: usa o teu visa ou mastercard para comprar Banana Gun (BANANA) instantaneamente.Saldo: usa os fundos da tua conta HTX para transacionar sem problemas.Terceiros: adicionamos métodos de pagamento populares, como Google Pay e Apple Pay, para aumentar a conveniência.P2P: transaciona diretamente com outros utilizadores na HTX.Mercado de balcão (OTC): oferecemos serviços personalizados e taxas de câmbio competitivas para os traders.Passo 3: armazena teu Banana Gun (BANANA)Depois de comprar o teu Banana Gun (BANANA), armazena-o na tua conta HTX.Alternativamente, podes enviá-lo para outro lugar através de transferência blockchain ou usá-lo para transacionar outras criptomoedas.Passo 4: transaciona Banana Gun (BANANA)Transaciona facilmente Banana Gun (BANANA) no mercado à vista da HTX.Acede simplesmente à tua conta, seleciona o teu par de trading, executa as tuas transações e monitoriza em tempo real.Oferecemos uma experiência de fácil utilização tanto para principiantes como para traders experientes.

141 Visualizações TotaisPublicado em {updateTime}Atualizado em 2026.06.02

Como comprar BANANA

Discussões

Bem-vindo à Comunidade HTX. Aqui, pode manter-se informado sobre os mais recentes desenvolvimentos da plataforma e obter acesso a análises profissionais de mercado. As opiniões dos utilizadores sobre o preço de BANANA (BANANA) são apresentadas abaixo.

活动图片

Categorias popularesright-arrow

Etiquetas Popularesright-arrow