看与听 · 图像、Sora 边界与语音See and Hear · Images, the Sora Boundary, and Voice.
文字之外，订阅还给了你图像、语音，以及很多旧教程会写到的 Sora / 视频入口。三样的实用度差很远 —— 图像在「够用的配图、改图」上最稳；语音的 Live 正逐步换成能同时听和说的 GPT-Live-1 / mini；Sora web/app 已停用，视频不要再按旧入口教。这章替你把它们各自的窄场景圈出来。Beyond text, the subscription gives you images, voice, and the Sora / video entrance many older tutorials mention. Their usefulness varies widely — images are steadiest for good-enough illustrations and edits; Voice Live is gradually moving to GPT-Live-1 / mini, which can listen and speak at once; and Sora web/app is discontinued. This chapter circles the narrow scenario each is actually good for.
文字之外，ChatGPT 订阅还打开了图像、语音，以及很多旧教程会写到的视频 / Sora 入口。这里最容易写偏：它不是万能创作平台，而是表达入口。Beyond text, a ChatGPT subscription opens images, voice, and the video / Sora entrance many older tutorials mention. This chapter is easy to miswrite: it is not an all-purpose production platform; it is an expression entrance.本章判断：图像、语音、视频省的是表达成本，不一定省事实核验成本。视觉和声音能让你更快把想法说清楚，但不能替你决定版权、品牌、事实和最终发布风险。This chapter's judgment: images, voice, and video save expression cost, not necessarily verification cost. Visuals and voice help you express an idea faster, but they do not decide copyright, brand, facts, or publishing risk for you.
— I
图像：先拿来做草稿Images: Use Them for Drafts First.
官方帮助说明里，ChatGPT Images 可以按文字生成图像，也可以上传现有图像后描述修改；它能按指令加文字、细节和透明背景，复杂图像最多可能需要约两分钟。1注 1Note 1OpenAI 帮助中心 · Creating images in ChatGPT —— 可按文字生成图像，也可上传图像后按描述编辑；可加文字、细节和透明背景；复杂图像最多可能需要约两分钟。OpenAI Help · Creating images in ChatGPT — images can be generated from text or edited from an uploaded image; they can include text, details, and transparent backgrounds; complex images may take up to about two minutes.这让它很适合封面探索、草图、风格变体、概念表达、社媒配图初稿。边界同样清楚：精确品牌资产、严肃事实图表、法律商用前的版权判断、医疗 / 金融 / 技术流程图，不能靠它一次生成就交付。Official Help says ChatGPT Images can generate from text or edit uploaded images based on instructions; it can add text, details, and transparent backgrounds, and complex images may take up to about two minutes.1注 1Note 1OpenAI 帮助中心 · Creating images in ChatGPT —— 可按文字生成图像，也可上传图像后按描述编辑；可加文字、细节和透明背景；复杂图像最多可能需要约两分钟。OpenAI Help · Creating images in ChatGPT — images can be generated from text or edited from an uploaded image; they can include text, details, and transparent backgrounds; complex images may take up to about two minutes. That makes it good for cover exploration, sketches, style variants, concept expression, and first drafts of social visuals. The boundary is just as clear: precise brand assets, serious factual charts, legal commercial copyright decisions, and medical / financial / technical diagrams should not be delivered from one generation.
提示词Prompt图像简报Image brief
用途：<封面 / 插图 / 概念图 / 社媒图 / 草图>
尺寸比例：<16:9 / 1:1 / 竖版封面>
主体：<必须出现的人 / 物 / 场景>
风格：<照片 / 插画 / 极简 / 编辑感 / 产品渲染>，参考<描述风格，不模仿在世艺术家>
文字：<必须出现的文字；如果必须精确，请单独说明>
避免：不要<品牌 logo / 真人肖像 / 错误图表 / 过度发光>
输出约束：给我 3 个方向，每个说明适用场景和风险。Use: <cover / illustration / concept image / social graphic / sketch>
Size and ratio: <16:9 / 1:1 / vertical cover>
Subject: <people / objects / scene that must appear>
Style: <photo / illustration / minimal / editorial / product render>, referencing <describe style without copying a living artist>
Text: <text that must appear; separately state if text must be exact>
Avoid: no <brand logo / real person / incorrect charts / excessive glow>
Output constraint: give me 3 directions, each with best use case and possible risk.
用途：技术博客的文章头图。
尺寸比例：16:9。
主体：一个抽象的「数据管道」概念——几条线汇聚到一个节点。
风格：极简扁平插画，编辑感，冷色调为主，参考瑞士国际主义平面设计的几何感。
文字：不需要文字。
避免：不要品牌 logo、不要真人肖像、不要过度发光。
输出约束：给我 3 个方向，每个说明适用场景和风险。Use: header image for a technical blog post.
Size and ratio: 16:9.
Subject: an abstract 'data pipeline' concept — a few lines converging into one node.
Style: minimal flat illustration, editorial feel, mostly cool tones, referencing the geometric sensibility of Swiss International Typographic design.
Text: no text needed.
Avoid: no brand logo, no real person, no excessive glow.
Output constraint: give me 3 directions, each with best use case and possible risk.
— II
视频：不要再把 Sora 当稳定入口教Video: Do Not Teach Sora as a Stable Entrance.
截至 2026-07-10，不能把 Sora web / app 当作 ChatGPT 订阅内稳定视频产品来教。As of 2026-07-10, Sora web / app cannot be taught as a stable video product inside the ChatGPT subscription.官方 Sora discontinuation 文档写明：Sora web 和 app experience 已于 2026-04-26 停用；Sora API 将于 2026-09-24 停用；已购 Sora credits 可转用于 Codex。2注 2Note 2OpenAI 帮助中心 · What to know about the Sora discontinuation —— Sora web 和 app experience 已于 2026-04-26 停用；Sora API 将于 2026-09-24 停用；已购 Sora credits 可转用于 Codex。OpenAI Help · What to know about the Sora discontinuation — the Sora web and app experiences were discontinued on 2026-04-26; the Sora API will be discontinued on 2026-09-24; purchased Sora credits can still be used for Codex.这不是小改名，而是产品状态变化。所以本书不写「如何用 Sora 做视频教程」。如果你看到旧教程让你去 sora.chatgpt.com 生成视频，要先按停用文档重新判断。The official Sora discontinuation article states that the Sora web and app experiences were discontinued on 2026-04-26; the Sora API will be discontinued on 2026-09-24; purchased Sora credits can be used for Codex.2注 2Note 2OpenAI 帮助中心 · What to know about the Sora discontinuation —— Sora web 和 app experience 已于 2026-04-26 停用；Sora API 将于 2026-09-24 停用；已购 Sora credits 可转用于 Codex。OpenAI Help · What to know about the Sora discontinuation — the Sora web and app experiences were discontinued on 2026-04-26; the Sora API will be discontinued on 2026-09-24; purchased Sora credits can still be used for Codex. This is not a minor rename; it is a product-status change. So this book does not teach "how to make videos with Sora." If an older tutorial sends you to sora.chatgpt.com for video generation, re-evaluate it against the discontinuation document first.
— III
语音：手不空时才值Voice: Valuable When Your Hands Are Busy.
Voice 现在要分三种：Live 是新一代自然对话，付费计划用 GPT-Live-1、Free 用 GPT-Live-1 mini；它能同时听和说，所以打断与接话更自然。Advanced 保留视频 / 屏幕共享等符合资格的移动能力，Standard 则是先转写再按轮次回答。Live 正在逐步推出，首发不进 Business / Enterprise / Edu、Temporary Chats、桌面 app、Work、Codex 与 custom GPTs，也暂不支持视频或屏幕共享。3注 3Note 3OpenAI 帮助中心 · ChatGPT Voice —— 截至 2026-07-10，Live 正在消费计划逐步推出：付费计划由 GPT-Live-1 驱动，Free 由 GPT-Live-1 mini 驱动；Live 可同时听和说。首发不支持 Business / Enterprise / Edu、Temporary Chats、桌面 app、Work、Codex、custom GPTs、视频或屏幕共享；可用性仍随计划、区域和版本变化。OpenAI Help · ChatGPT Voice — as of 2026-07-10, Live is rolling out across consumer plans: paid plans use GPT-Live-1 and Free uses GPT-Live-1 mini; Live can listen and speak at the same time. At launch it does not support Business / Enterprise / Edu, Temporary Chats, the desktop app, Work, Codex, custom GPTs, video, or screen sharing; availability still varies by plan, region, and version.所以「已驱动 ChatGPT Voice」不能写成「所有账号、所有表面都已切换」。Voice now has three lanes. Live is the new natural conversation experience, powered by GPT-Live-1 on paid plans and GPT-Live-1 mini on Free; it can listen and speak at once, making interruption and turn-taking more natural. Advanced retains eligible mobile capabilities such as video and screen sharing, while Standard transcribes first and answers turn by turn. Live is rolling out and, at launch, is not in Business / Enterprise / Edu, Temporary Chats, the desktop app, Work, Codex, or custom GPTs; it also lacks video and screen sharing.3注 3Note 3OpenAI 帮助中心 · ChatGPT Voice —— 截至 2026-07-10，Live 正在消费计划逐步推出：付费计划由 GPT-Live-1 驱动，Free 由 GPT-Live-1 mini 驱动；Live 可同时听和说。首发不支持 Business / Enterprise / Edu、Temporary Chats、桌面 app、Work、Codex、custom GPTs、视频或屏幕共享；可用性仍随计划、区域和版本变化。OpenAI Help · ChatGPT Voice — as of 2026-07-10, Live is rolling out across consumer plans: paid plans use GPT-Live-1 and Free uses GPT-Live-1 mini; Live can listen and speak at the same time. At launch it does not support Business / Enterprise / Edu, Temporary Chats, the desktop app, Work, Codex, custom GPTs, video, or screen sharing; availability still varies by plan, region, and version. So "powers ChatGPT Voice" must not become "already switched for every account and surface."它的甜区没有因此变成「替你做严肃研究」。更自然的双向语音适合口语练习、头脑风暴、走路时梳理思路、开车前的轻量计划、面试 / 演讲陪练；仍不适合精确引用、严肃事实核验、复杂表格，语音转写也不是逐字记录。That does not turn it into a serious-research tool. More natural duplex voice is useful for speaking practice, brainstorming, walking through an idea, lightweight planning before driving, and interview / presentation rehearsal. It still does not fit precise citations, serious verification, or complex tables, and its transcript is not verbatim.
需求Need
用什么Use
别用什么Avoid
够用的配图、草稿、风格方向Good-enough visual, draft, style direction