CREATE WITH AI

AI 创作与商业情报

AI for creative work and business.

关注能用于创作、营销与电商的产品更新，附原始来源。

Product updates for creation, marketing, and commerce, with original sources.

2026.09.09 · LATEST最新一期日报Latest daily issue阅读日报 →Read daily →

最新情报

Latest intelligence

10
AI CreationOpenAI2026-09-08

官方发布

Official update

ChatGPT Images 2.5：局部修改与多轮编辑进入新一轮升级ChatGPT Images 2.5 improves targeted and multi-turn editing

OpenAI 9 月 8 日发布 Images 2.5，并推出 API 模型 Flare 与 Sunburst。官方强调参考主体保留、局部修改和连续编辑；Sunburst 面向更精细的成稿，生成等待更长。

OpenAI announced Images 2.5 on September 8, alongside Flare and Sunburst in the API. It reports better subject preservation and more reliable targeted and successive edits. Sunburst targets detailed final assets with longer generation times.

编辑观察：商品图团队可用同一张实拍，依次改背景、文案和构图，逐轮检查包装文字、Logo 与形状。记录修改次数和最终可用率，再决定是否切换批量流程。

Editorial note: Test successive background, copy and layout edits on one product photo. Check packaging, logos and shape after each turn; record revision count and usable-output rate before changing a production pipeline.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CreationAdobe2026-09-08

官方发布 · 含公测

Official update

Premiere 与 After Effects：生成、声音修复和项目助手一起进入剪辑现场Adobe brings generation, audio cleanup and assistance into the edit

Adobe 9 月 8 日介绍 Premiere 的 Generative Media Tool，可在时间线上生成视频和音效，并选择 Firefly 或合作模型。After Effects 的 AI Assistant 进入公开测试，支持项目整理与多步骤操作。

Adobe’s September 8 update introduces timeline video and sound generation in Premiere, with Firefly and partner-model choices. After Effects gains an AI Assistant in public beta for project organization and multi-step tasks.

编辑观察：先复制一份短片工程，分别测试补镜头、访谈串音和 AE 图层整理。验收是否保留可编辑结构、是否能撤回，以及原项目能否继续由同事接手。

Editorial note: Use a duplicate project to test missing shots, overlapping speech and AE layer organization. Check editability, reversibility and handoff to another editor.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CreationRunway2026-09-08

官方发布 · 付费生成

Official update

Runway 插件进入 Premiere / AE，减少导出再导入的往返Runway’s Adobe plugin closes the export–generate–reimport loop

Runway 9 月 8 日发布 Adobe 插件，在 Premiere Pro 与 After Effects 面板内生成图像、视频并把结果送回序列或合成。Edit Studio 可用 Aleph 2 重绘现有片段。

Runway’s September 8 plugin places image and video generation inside Premiere Pro and After Effects. Results return to the sequence or composition; Edit Studio uses Aleph 2 to transform existing clips.

编辑观察：优先试用你已有的一个片段：核对源片与输出的时长、帧率、色彩和费用。若时间线衔接稳定，再把插件纳入剪辑规范。

Editorial note: Start with an existing clip and check duration, frame rate, color and credit use. Adopt it only after timeline handoff is dependable.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CommerceMeta2026-09-08

官方发布 · 美国分批开放

Official update

Muse 能代办购物：Link 已接入，Shop Pay 仍待上线Muse brings agent checkout through Link; Shop Pay is still pending

Meta 9 月 8 日发布 Muse，可在专用云端电脑中浏览网站、填写表单和处理任务。官方明确写到 Link 结账；发送邮件或购买前会请求确认，因此它与电商工作流有直接关系。

Meta’s September 8 Muse announcement explicitly includes browsing, form filling and checkout through Link. It requests approval before purchases or emails, making this directly relevant to commerce workflows.

编辑观察：商家可检查商品规格、库存、运费和退货条件是否清楚且一致，供代理读取；这是一项页面准备工作，尚不能据此判断代理会带来多少订单。

Editorial note: Audit consistency of product specifications, stock, shipping and returns for agent readers. This is a readiness exercise, not evidence of incremental orders.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI MarketingElevenLabs2026-09-08

新指南 · 非模型发布

Guide / operational context

声音也需要品牌规范：把配音、片头声与客服语气连起来Treat voice and sonic identity as reusable brand assets

ElevenLabs 的声音品牌指南初发于 9 月 8 日，页面另标 9 月 9 日更新。它讨论跨触点的声音一致性，属于方法内容，而非新的语音模型或已证实的广告效果提升。

ElevenLabs’ sonic-branding guide is dated September 8 and updated September 9. It discusses audio consistency across brand touchpoints; it is guidance, not a model launch or proof of advertising lift.

编辑观察：写一页声音规范：语速、情绪、发音表、禁用语气、语言版本与授权记录。用一条 15 秒广告和一段客服开场检查是否像同一个品牌。

Editorial note: Document pace, emotion, pronunciation, prohibited tones, language variants and permissions. Compare a short ad with a support greeting for brand consistency.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CreationArena2026-09-07

榜单观察 · 非发布日

Leaderboard snapshot

Arena 图像编辑榜：Sunburst / Flare 在前列，仍是 PreliminaryArena image-edit snapshot: Sunburst and Flare remain preliminary

9 月 9 日读取的页面标注 9 月 7 日，覆盖 55 个模型、29,526,046 票。Sunburst 为 1520±9（6,704 票），Flare 为 1491±9（5,676 票），两者均标 Preliminary。

The page retrieved on September 9 is dated September 7: 55 models and 29,526,046 votes. Sunburst scores 1520±9 from 6,704 votes; Flare scores 1491±9 from 5,676. Both are preliminary.

编辑观察：把榜单用于选出对照模型，再用同一套商品图做盲测。至少包含中文标签、透明包装、局部换色、保留 Logo 和连续修改，逐项记失败类型。

Editorial note: Use the ranking to shortlist models, then blind-test identical product assets for Chinese labels, transparent packaging, local recoloring, logo retention and repeated edits.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CreationGoogle2026-09-04

近期补读／本月提醒 · 原始日期见左

Catch-up / context · original date shown

Lyria 3.5 进入 Gemini 和 API，配乐工作流多了直接入口Lyria 3.5 reaches Gemini and the API

Google 9 月 4 日宣布 Lyria 3.5 在 Gemini 应用和 Gemini API 可用，提供人声／纯音乐选择、模板和长短曲目选项。这是入口扩展；模型此前已于 7 月进入 Flow Music。

Google announced Lyria 3.5 in Gemini and the Gemini API on September 4, adding vocal/instrumental choices, templates and track-length options. This is an access expansion after its July Flow Music release.

编辑观察：为同一短片分别做纯音乐与人声版本，比较节奏切点、对白遮盖和可循环性；投放前按实际产品条款核对使用范围。

Editorial note: Compare instrumental and vocal versions against the same edit for timing, dialogue masking and loopability; verify the applicable terms before campaign use.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CreationRunway2026-09-04

近期补读／本月提醒 · 原始日期见左

Catch-up / context · original date shown

Runway Team：共享素材、品牌资料与额度进入小团队方案Runway Team pools projects, brand assets and generation credits

Runway 9 月 4 日推出面向 2–9 席的小团队方案。每席每月带来 6,900 credits，进入共同额度池；共享项目、Brand Kits 和评论让多人生产能沿用同一套参考资料。

Runway launched Team on September 4 for 2–9 seats. Each seat contributes 6,900 monthly credits to a shared pool, with shared projects, Brand Kits and comments.

编辑观察：团队先盘点实际编辑人数、每月可用产出与返工消耗，再评估共享池。把参考图和品牌资料统一，避免只增订阅席位却没有改善交接。

Editorial note: Audit editors, usable outputs and revision costs before migrating. Consolidate references and brand assets so shared access improves handoff.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI CommerceElevenLabs / Allegro2026-09-07

近期补读／本月提醒 · 原始日期见左

Catch-up / context · original date shown

Allegro 的物流语音客服：先接住高频问题，再带着上下文转人工Allegro’s voice support focuses on delivery questions and contextual handoff

ElevenLabs 当前案例页显示 9 月 7 日，介绍 Allegro 在三条物流热线部署语音代理，处理包裹进度等高频问题；需要人工时传递运单号和对话摘要。

ElevenLabs’ current case page, dated September 7, describes a voice-agent rollout across three Allegro delivery hotlines, with tracking numbers and summaries handed to human staff.

编辑观察：客服团队可先选一个问题量大、答案明确的队列，记录转人工率、问题解决率和错误回答。交接时保留客户已经提供的信息，避免让对方重新讲一遍。

Editorial note: Pilot a narrow, high-volume queue and track handoff, resolution and incorrect answers. Preserve context so customers do not repeat themselves.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
AI MarketingGoogle Ads2026-06-11

近期补读／本月提醒 · 原始日期见左

Catch-up / context · original date shown

AI Max 迁移别看旧摘要：ACA 本月开始，DSA 已延至 2027 年AI Max migration: September for ACA, February 2027 for DSA

Google 原文 4 月发布、6 月 11 日修订：DSA 的自动迁移开始时间已延至 2027 年 2 月；ACA 与 campaign-level broad match 仍从 2026 年 9 月开始。旧摘要仍可能把三者混在一起。

Google’s April announcement was amended on June 11: DSA migration now starts in February 2027, while ACA and campaign-level broad match still start in September 2026.

编辑观察：先查账户通知，再登记受影响活动、匹配设置和文本定制状态。试验期间保留对照组与修改记录；不要因读到自动迁移就同时大改预算、素材和落地页。

Editorial note: Check account notices, inventory affected campaigns and preserve a control group and change log. Avoid changing budget, creative and landing pages simultaneously.

详情、范围与来源Details, scope and sources原始来源 / Source ↗
2026-W36 · 2026.08.31—09.06
WEEKLY

从模型发布到实际交付

From model release to usable work

用一件商品，验证资料、卖点、分镜与页面能否连成一次合格交付。

Use one product to test research, copy, storyboarding, and page delivery.

阅读专题 →Read feature →