AI NEWSLatest AI NewsArticleXiaopeng Robotaxi Begins Unmanned Road Testing, Equipped with Second-Generation VLA Has Completed Over 2000 Internal TestsPublished in Latest AI NewsTime :Aug 27, 2026Read :3minuteXiaopeng Auto announced on August 27 that its Xiaopeng Robotaxi, equipped with the second-generation VLA large model, has officially started fully driverless road testing. It has also obtained the remote testing qualification for intelligent connected vehicles in Guangzhou and can conduct autonomous driving tests on relevant first-, second-, and third-level test roads in the city.According to the introduction, since the employee internal testing of Xiaopeng Robotaxi started in June this year, it has completed more than 2,000 internal test orders within two months, verifying the complete process from online order placement, vehicle pickup, to trip services. (17173News) Previously, Xiaopeng has been advancing product development, testing, and operation system construction for the Robotaxi business and plans to further promote driverless passenger operations.The second-generation VLA large model equipped in this Robotaxi is a multimodal AI model developed by Xiaopeng for the physical world, integrating visual, language, and action decision-making capabilities. It can be used in embodied intelligence scenarios such as autonomous driving, Robotaxi, and humanoid robots. This model achieves on-vehicle deployment of large models through collaborative optimization of “chip—compiler—model” and improves perception and decision-making capabilities in complex environments.As autonomous driving technology evolves from driver assistance to high-level intelligent driving, Robotaxi has become an important direction for car manufacturers to explore AI commercialization. Xiaopeng’s advancement in driverless testing indicates that its autonomous driving technology based on large models is entering a larger-scale road verification stage, which will further promote the application exploration of physical AI in the transportation field.

AI NEWSLatest AI NewsArticleNVIDIA Acquires Hugging Face for $1.29 Billion, Open-Source AI Foundation Joins the Computing EmpirePublished in Latest AI NewsTime :Aug 27, 2026Read :2minuteAccording to The Information, NVIDIA is set to acquire the open-source AI platform Hugging Face for $12.9 billion. The company, which hosts open-source models and provides cloud services, has led the market for a decade, with annual revenue of about $150 million and nearing profitability. NVIDIA’s offer is approximately 80 times its annual revenue.“Closed-Source Labs” Decoupling Forces NVIDIA to Compete for EcosystemNVIDIA had previously committed to investing $26 billion over five years to develop open-source AI models. However, closed-source companies such as OpenAI and Anthropic are actively reducing their reliance on NVIDIA hardware. OpenAI has partnered with Broadcom to develop its own chips, signed a $1 billion agreement with Cerebras, and secured AMD production capacity. Anthropic has also been advancing its own chip development, alongside Google. The trend towards self-sufficient computing power is evident.By acquiring Hugging Face, NVIDIA can expand its cloud business and position it as “OpenRouter,” an open-weight model platform. This aligns with its existing strategy and leverages developer reputation to find new markets for its own chips. At a time when closed-source labs are retreating, this acquisition reclaims control of the open-source ecosystem for the king of computing power.

AI NEWSLatest AI NewsArticleCCTV Exposes AI Dating Scam: Fraudulent Matchmaking Targets Middle-Aged and Older Women, 28 People CaughtPublished in Latest AI NewsTime :Aug 27, 2026Read :3minuteAccording to CCTV News, a recent AI dating fraud targeting middle-aged and elderly women has emerged. Unlawful individuals use AI-generated fake dating partners to defraud money. The Chengdu police in Sichuan recently dismantled an “AI dating” gang, arresting 28 suspects including Wang, with the case involving over one million yuan and affecting more than 50 victims.“Two-Act Play”: AI Image + Real Person After-sales PerformanceComplainant Ms. Bian saw a matchmaking company on a video platform, and someone with the nickname “Director Wang of the Matching Center” sent her a profile and photo of “Li Ge.” She paid 2000 yuan for the service but was never able to contact the person. Police investigation found that the signing company did not exist, but only sales, editing, and host positions were present at the site. The editing computer contained the same AI male guest image that Ms. Bian had received.The gang used AI male images for live broadcasts and pre-written scripts, with hosts using voice changers to impersonate “high-quality males” for live interactions. After-sales personnel then continued the performance with victims, never offering real dating. Although Wang returned Ms. Bian’s money, the police seized conclusive evidence by monitoring his live broadcast account and arrested him while he was still using AI images for live streaming.CCTV pointed out that such fictitious identity fraud is evident and still spreading on short video platforms, calling on platforms to establish risk classification handling mechanisms and transfer fraud clues to relevant authorities. Experts advocate for implementing AI-generated content labels, improving identification risk control, and blocking推送 from algorithms; ordinary users can also use technology to verify the authenticity of suspected AI content and strengthen risk awareness.

Join NowENLatest AI NewsCCTV Exposes AI Dating Scam: Fraudulent Matchmaking Targets Middle-Aged and Older Women, 28 People CaughtCCTV reports AI matchmaking scams targeting middle-aged and elderly women. Chengdu police busted a gang of 28, involving over 1 million yuan and about 50 victims. Scammers used AI-generated images with human after-sales; one victim paid 2,000 yuan but couldn’t reach them…..6 hours ago13.0KNVIDIA Acquires Hugging Face for $1.29 Billion, Open-Source AI Foundation Joins the Computing EmpireNvidia plans to acquire Hugging Face for $12.9B. Hugging Face has ~$150M annual revenue and is near profitability; the bid is ~80x revenue. Closed-source labs are reducing reliance on Nvidia hardware; OpenAI partners with Broadcom on custom chips. Nvidia pledged $26B over 5 years for open-source models to capture the ecosystem…..7 hours ago11.8KXiaopeng Robotaxi Begins Unmanned Road Testing, Equipped with Second-Generation VLA Has Completed Over 2000 Internal TestsOn Aug 27, XPeng announced its Robotaxi with the second-gen VLA model started fully driverless road tests after receiving Guangzhou remote testing qualification. Since June employee beta, 2,000+ orders were completed in two months, validating booking, pickup, and trip services…..7 hours ago11.3KPublic Welfare Advertisement: Pedestrian Head and Feet Disordered, Garbage Bags Emerge from Brooms - Official Confirms It Was AI-Generated and Has Been RemovedA public service ad in Linfen, Shanxi was suspected AI-generated due to distorted images: pedestrians with mismatched heads and feet, a broom emerging from a garbage bag. After exposure, local officials replaced it overnight. The government confirmed the ad was AI-made by a partner agency; review only checked text, not visuals…..7 hours ago10.9KCompressed from 3.3GB to 440MB! How does Tencent Hunyuan’s ultra-low bit translation model achieve near-zero loss?Tencent Hunyuan compressed a 1.8B edge-side translation model to hundreds of megabytes, achieving near-zero loss in translation quality, and has been applied to real-time translation of B站 live chat messages, supporting high-concurrency real-world applications. The model supports mutual translation among 33 languages, and its translation quality leads among models of the same size.8 hours ago11.3KClaude Desktop Cowork Built-in Browser Released, Can Complete Web Operations IndependentlyClaude desktop Cowork adds a built-in browser that auto-opens in the sidebar for web tasks. It can browse, read, click, and type autonomously without extensions or setup. It is fully isolated from your daily browser, seeing no tabs, bookmarks, or saved passwords; handle site logins separately if needed…..8 hours ago10.8KAmazon Orders Triple, NVIDIA Chips Become Global Big Tech Hard CurrencyAmazon and Nvidia deepen ties: after deploying over 1M GPUs, orders tripled, planning 2M more, including Blackwell Ultra and Rubin chips, underscoring surging AI demand and the huge capital race behind computing power…..8 hours ago11.5KBeijing Imposes First AIGC Business Secret Administrative Penalty Case, Company Fined 100,000 for Taking AI Software from Previous CompanyBeijing Chaoyang’s market regulator fined 100,000 yuan in the city’s first AIGC trade secret administrative penalty case for unauthorized copying and use of a former employer’s AI software, setting compliance red lines for the digital economy. The rights holder specializes in foreign-language dubbing for micro-dramas and AI multilingual film translation…..9 hours ago10.5KGoogle Upgrades Gemini Live to Integrate with Spark, Voice Controls Gmail and Entire SuiteGoogle upgraded Gemini Live from a voice conversation tool toward productivity and intelligent agent workflows. Supporting 40+ languages since 2024 and deep Google app integration, the update strengthens task execution, cross-app collaboration, and personalization…..9 hours ago11.9KQwen3.8-Flash Agent Officially Launched, Qwen Office Enters the Era of Abundant and Efficient ManagementQianwen Office Platform launches Qwen3.8-Flash and Standard mode for all users, with lower credit use and faster token throughput. A dual-model system will follow: Standard mode covers 95% of daily office tasks, Advanced mode handles the remaining 5% complex tasks…..9 hours ago11.8KTaking AI Software to Start a Business: Beijing’s First AIGC Trade Secret Case Fines 100,000 YuanBeijing Chaoyang market regulator fined 100,000 yuan in the city’s first AIGC trade secret administrative penalty, involving an employee taking secrets to a new job and copying/using the former employer’s AI software. It formed a full evidence chain of copy-crack-reuse across firms; the rights holder is a key exporter of micro-drama dubbing and AI multilingual film translation…..10 hours ago12.4KMade $118 Billion in Half a Year! NVIDIA Releases Explosive Financial Report, Next-Generation AI Chips Go into Mass ProductionNVIDIA released its half-year report and second-quar

Join NowENPremium Membership ·Limited-Time Pricing— Save TodayChoose a monthly plan that fits your needs. Each tier includes tailored points allowances and flexible usage limits for confident monitoring.Your current plan isFreeEssentialStart with essential GEO monitoring and analysis$14.9/30 daysGet PremiumPoints IncludedOne-time points 7,000Daily bonus points 100AI Query Mining50 usesGEO Rank Check50 uses5 AI platforms:GEO Brand Score Audit3 uses5 AI platforms:GEO Link Citation Tracking5 uses5 AI platforms:GEO Rank TrackingNot includedData Export：Not includedReport Generation：Not included7-Day Featured Slot in AI DirectoryNot includedDedicated SupportNot includedMost PopularGrowthBuilt for growing brands and daily GEO operations$45/30 daysGet PremiumPoints IncludedOne-time points 25,000Daily bonus points 200AI Query Mining150 usesGEO Rank Check100 uses7 AI platforms:GEO Brand Score Audit8 uses7 AI platforms:GEO Link Citation Tracking15 uses7 AI platforms:GEO Rank TrackingNot includedData Export：Not includedReport Generation：Not included7-Day Featured Slot in AI DirectoryNot includedDedicated SupportNot includedProProfessional GEO monitoring and continuous growth analysis$109/30 daysGet PremiumPoints IncludedOne-time points 65,000Daily bonus points 300AI Query Mining300 usesGEO Rank Check300 uses10 AI platforms:GEO Brand Score Audit20 uses10 AI platforms:GEO Link Citation Tracking30 uses10 AI platforms:GEO Rank TrackingUnlimited10 AI platforms:Data Export：IncludedReport Generation：Included7-Day Featured Slot in AI DirectoryNot includedDedicated SupportNot includedEnterprise PlanEnterpriseScalable GEO operations for enterprises and agencies$199/30 daysGet PremiumPoints IncludedOne-time points 120,000Daily bonus points 500AI Query Mining800 usesGEO Rank Check600 uses10 AI platforms:GEO Brand Score Audit30 uses10 AI platforms:GEO Link Citation Tracking60 uses10 AI platforms:GEO Rank TrackingUnlimited10 AI platforms:Data Export：IncludedReport Generation：Included7-Day Featured Slot in AI Directory1 usesDedicated SupportIncludedCompare PlansCompare usage allowances and premium benefits across plans at a glanceFree$0Get PremiumEssential$99/30 daysGet PremiumGrowth$299/30 daysGet PremiumPro$699/30 daysGet PremiumEnterprise$1,299/30 daysGet PremiumPoints IncludedOne-time points07,00025,00065,000120,000Daily bonus points100100200300500AI Query MiningQuery allowance550150300800GEO Rank CheckQuery allowance350100300600AI platformsGEO Brand Score AuditQuery allowance1382030AI platformsGEO Link Citation TrackingQuery allowance15153060AI platformsGEO Rank TrackingGEO Rank TrackingNot includedNot includedNot includedUnlimitedUnlimitedData ExportNot includedNot includedNot includedIncludedIncludedReport GenerationNot includedNot includedNot includedIncludedIncluded7-Day Featured Slot in AI Directory7-Day Featured Slot in AI DirectoryNot includedNot includedNot includedNot included1Dedicated SupportDedicated SupportNot includedNot includedNot includedNot includedIncludedFrequently Asked QuestionsCommon questions about Aibase memberships, credits, GEO monitoring, and billingWhat is the difference between monthly membership credits and daily login bonus credits?Monthly membership credits are the core AI computing allowance included with your membership and are issued once per membership benefit cycle.Daily login bonus credits are an additional benefit. They are automatically issued when a member first signs in to or visits Aibase on a given day, with no manual claim required.Daily bonus credits are valid only for the day they are issued and do not roll over to the next day.Do I need to manually claim daily bonus credits? Are missed days credited later?No manual claim is required. Daily bonus credits are automatically issued when you first sign in to or visit Aibase that day.If you do not sign in to or visit Aibase on a given day, that day’s bonus credits will not be issued and will not be credited retroactively.Daily bonus credits are an additional membership benefit and do not affect your regular monthly membership credits.Do monthly membership credits roll over to the next month?No. Membership credits are issued in 30-day benefit cycles.Unused membership credits expire at the end of the current benefit cycle and do not roll over into the next cycle.At the beginning of a new benefit cycle, the credit allowance for your current membership plan is issued again.If I buy a 3-, 6-, or 9-month membership, will all credits and usage quotas be issued at once?No. A multi-month purchase provides membership access for the selected duration; it does not issue all future credits and usage quotas upfront.Membership benefits are still managed in 30-day cycles. Credits are issued and tool usage quotas are reset at the start of each cycle.For example, a 3-month membership consists of three consecutive 30-day benefit cycles.In what order are different credit balances used?Credits are deducted in the following order: daily login bonus credits →

Play with Putty – 谷歌推出的实时协作 Vibe Coding 工具AI工具1天前更新AI小集02Play with Putty是什么Play with Putty 是 Google Labs 推出的多人实时协作 vibe coding 实验工具。用户用自然语言描述需求，AI 自动生成代码，多人可在同一共享空间实时搭建工具与网站，修改即时同步，无需编程基础。产品目前采用候补名单制分批开放，定位轻量社交化创作，让软件从个人编写变为集体对话，在实时协作中快速成型。Play with Putty的主要功能多人实时协作：支持多人进入同一共享工作空间，像编辑协作文档一样同时搭建项目。自然语言驱动（Vibe Coding）：用户用自然语言描述需求，AI 自动生成对应代码，无需编程基础。实时同步渲染：任何提示词修改即时同步给所有参与者，无需手动保存或刷新页面。生成可用产品：输出结果是可直接使用的工具或网站，而非静态预览，支持即时测试。链接邀请机制：创建空间后通过分享链接邀请他人加入，快速组建协作团队。轻量易用：定位消费级实验工具，比 Google AI Studio 更轻量，开箱即用。如何使用Play with Putty访问官网：访问 Play with Putty 官网 https://labs.google/playwithputty，用个人 Google 账户登录并加入 waitlist。等待邀请：Google 会分批通过邮件发送访问邀请，Workspace 及学校/企业账户暂不支持。创建共享空间：获得权限后，新建一个 Putty 空间，系统会自动生成邀请链接。邀请协作者：将链接分享给其他人，对方点击即可加入同一实时协作房间。自然语言描述需求：在共享空间中用自然语言输入提示词，告诉 AI 你想搭建什么工具或网站。实时查看生成结果：AI 自动生成代码并即时渲染，所有参与者同步看到更新，无需保存或刷新。迭代修改优化：房间内任何人都可以随时修改提示词，AI 会立即重新生成，团队可边讨论边搭建。Play with Putty的核心优势多人实时协作：支持多人进入同一共享空间同时创作，打破传统单人编码的孤立体验。零代码门槛：用户只需用自然语言描述需求，AI 自动生成代码，无需任何编程基础。实时同步更新：所有修改即时同步给房间内每位参与者，无需保存或刷新即可看到最新效果。产出即产品：生成结果不是静态预览，而是可直接使用的工具或网站，支持即时测试验证。对话式创作：将软件开发变成集体头脑风暴，团队在讨论中实时看到想法落地，大幅缩短迭代周期。轻量易上手：作为消费级实验工具，无需复杂配置，开箱即用，比专业开发平台门槛更低。Play with Putty的同类竞品对比对比维度Play with PuttyBolt.new产品定位多人实时协作 vibe coding 实验工具单人 AI 全栈应用构建与部署平台协作模式核心卖点：多人同空间实时编辑，共享光标、即时同步以个人开发为主，支持分享链接但非实时协同编辑交互方式自然语言提示词驱动，对话式集体创作自然语言提示词 + 代码编辑器，个人与 AI 结对编程输出形态生成可即时测试的工具或网站，侧重快速验证生成可一键部署的生产级全栈应用（含数据库、后端）目标用户非技术人员、创意团队、教育工作者、轻量用户开发者、独立黑客、希望快速上线的技术人员使用门槛零代码，消费级轻量体验低代码，但理解代码结构有助于深度定制部署能力实验性质，暂未强调独立部署支持一键部署至生产环境，自带域名与托管Play with Putty的应用场景团队快速原型验证：产品、设计、开发多人同屏，用自然语言边讨论边生成可交互原型，1 小时内将想法变为可点击 demo。编程教学与课堂互动：老师与学生共处一个空间，学生用语言描述需求，AI 实时生成代码，老师即时讲解逻辑，降低入门门槛。黑客马拉松与创意马拉松：队员无需配置环境，通过链接加入即可集体 vibe coding，快速拼凑 MVP 并现场演示。企业内部小工具搭建：运营、销售等非技术同事协作搭建数据看板、计算器、审批表单等轻量工具，即搭即用。朋友聚会娱乐创作： 一群朋友边语音边协作，用语言描述一起做个小游戏、投票页面或纪念网站，将社交与创作合二为一。# AI工具©版权声明本站文章版权归AI工具集所有，未经允许禁止任何形式的转载。上一篇AZ8 - AI 视频创作工作台，Agent 实时读取素材与状态下一篇Omarchy - 开源的AI原生Linux桌面 ，自然语言生成桌面主题相关文章AnimateDiff-Lightning – 字节推出的快速生成高质量视频的模型AI小集3Pixtral Large – Mistral AI开源的超大多模态模型AI小集3ST-Raptor – AI表格问答工具，支持多种半结构化表格AI小集3scite – AI文献检索与分析工具，智能分析引文和自动分类AI小集2LONGSHOT – AI内容创作与SEO优化助手，根据实时数据和行业趋势生成创意内容AI小集2IQuest-Coder-V1 – IQuestLab推出的代码大模型系列AI小集2暂无评论再想想发表评论暂无评论…热门工具Loomy即梦AISekoAiPPT秘塔AI搜索妙呀堆友Agent美图设计室绘蛙AIupdream办公小浣熊秒哒最新收录MetaDigChatArtLexcat造剧立刻成片OJO最新文章秒悟Meoo – 1分钟生成前后端完整的网站，真0门槛开发！25分钟前怎么用 MetaDig 做 AI 短视频，附教程实测案例27分钟前Gemini 3.5 Transcribe – 谷歌推出的最新语音转文本模型4小时前Hy-MT2-1.8B – 腾讯混元推出的端侧翻译大模型4小时前GLM-5.3-Flash – 智谱开源的原生多模态模型，即Ox Alpha10小时前Qwen3.8-Flash – 阿里通义推出的多模态 MoE 模型10小时前Omarchy – 开源的AI原生Linux桌面 ，自然语言生成桌面主题10小时前AZ8 – AI 视频创作工作台，Agent 实时读取素材与状态1天前Tutti · VM – 多人多 Agent 并行协作空间，零延迟并行协作1天前EgoSuite-Open100K – 光轮智能开源的全模态人类行为数据集1天前PixVerse R2 – 爱诗科技推出的实时全模态世界模型1天前Breeze TTS 2 – BreezeBlue 推出的新一代语音合成模型1天前QuickDesk – 开源 AI 原生远程桌面，首款内置 MCP Server1天前AlphaVow – 一站式 AI 批量图像处理工具2天前Agnes 2.5 Pro Alpha – Agnes AI 开源的多模态推理模型2天前

IT之家8 月 26 日消息，Stability AI 美国加州当地时间 25 日宣布完成总额 7,600 万美元（IT之家注：现汇率约合 5.12 亿元人民币）的 B 轮融资。这笔资金将用于推动创意制作产品的开发，深化应用研究领域，扩展专业服务业务。Stability AI 本次融资的投资者包括知名游戏企业 EA、全球三大音乐娱乐企业（索尼音乐、环球音乐、华纳音乐），AMD 风投部门也在这一轮中出资。Stability AI 首席执行官 Prem Akkaraju 表示：这一无与伦比的投资者群体印证了我们愿景的实现 —— 即生成式 AI 能够赋能每一位制片人、音乐家和故事讲述者。Stability 在人工智能领域独树一帜，因为我们是为创意人士打造工具的富有创造力的人。开发具有行业引领性的技术意味着要聚集一群具有行业引领力的巨头，而这正是我们做到的。我们的投资者为我们提供了长期成功的保障，因为他们不仅提供资金，还提供专业知识、信誉以及与艺术家直接联系的机会。

PixVerse R1 – 爱诗科技推出的全球首个通用实时世界模型AI工具5个月前更新AI小集03PixVerse R1是什么PixVerse R1 是爱诗科技（PixVerse）推出的全球首个通用实时世界模型，标志着AI视频生成从“预录制”迈向“实时动态生成”的重大突破。R1通过三大核心技术实现创新：Omni原生多模态基础模型，将文本、图像、音频与视频融合为单一生成序列；自回归流式生成机制，解决长时序一致性问题，支持用户实时插入指令；瞬时响应引擎（IRE），大幅提升计算效率，实现即时响应。 访问PixVerse平台可直接体验PixVerse R1 的最新能力。PixVerse R1的主要功能实时视频生成：能即时生成连续的视频内容，支持最高1080P分辨率，用户可在生成过程中随时插入新指令，视频会即时响应并调整画面。自回归流式生成：解决了长时序一致性问题，即使在不干预的情况下，视频也能自主发展剧情，实现无限连续的视觉流式传输。多模态融合：采用Omni原生多模态基础模型，将文本、图像、音频与视频融合为单一生成序列，避免了传统上采样带来的模糊与伪影问题。高效计算引擎：通过瞬时响应引擎（IRE），大幅压缩采样步数，计算效率提升数百倍，实现低延迟的实时交互体验。数字分身系统：上传照片能生成三种风格（写实/动漫/赛博朋克）虚拟角色，可在不同场景中探索互动。多人互动直播：支持多人在同一虚拟空间实时创作，Prompt以弹幕形式触发画面变化，所有人可见。实时聊天：Live Chat功能支持创作者即时交流创意。PixVerse R1的技术原理Omni原生多模态基础模型：采用Transformer架构，将文本、图像、音频与视频融合为一个统一的生成序列，避免了传统方法中因上采样导致的模糊和伪影问题，提升了生成内容的质量和一致性。自回归流式生成机制：通过自回归的方式实现视频的连续生成，解决了长时序一致性问题。用户可以在生成过程中实时插入新的指令，系统能即时响应并动态调整视频内容，使生成的视频能够根据用户的输入灵活变化。瞬时响应引擎（IRE）：通过时间轨迹折叠、引导校正和自适应稀疏采样等技术，大幅压缩了传统扩散模型所需的采样步数，显著提升了计算效率，实现了低延迟的即时响应能力，为实时交互提供了技术支撑。实时交互与决策：PixVerse R1支持用户在视频生成过程中随时输入新的指令（Prompt），系统会根据这些指令即时调整视频的叙事方向，在没有用户干预时，模型能自主生成后续内容，实现“无限流”的视频生成体验。高效算力优化：针对实时生成的高算力需求，PixVerse R1通过优化算法和架构设计，降低了单次生成的资源消耗，使得长时间的实时视频生成成为可能，保持了高质量的生成效果。PixVerse R1的项目地址项目官网：https://pixverse.ai/en/blog/pixverse-r1-next-generation-real-time-world-modelPixVerse R1的应用场景实时互动娱乐：可用于创建实时互动的视频内容，如互动剧情游戏，观众可以通过输入指令即时改变故事走向，带来全新的沉浸式娱乐体验。直播与内容创作：支持实时生成视频内容，创作者可以根据观众反馈即时调整直播画面或视频剧情，为直播和短视频创作带来更多创意和互动性。教育与培训：在教育领域，可用于生成实时的虚拟场景或模拟环境，帮助学生更好地理解和参与学习内容，如历史事件重现、科学实验模拟等。广告与营销：品牌可以用PixVerse R1生成个性化的广告视频，根据用户行为或偏好实时调整广告内容，提升广告的吸引力和效果。游戏开发：为游戏开发者提供实时生成的游戏世界和剧情，实现更灵活的游戏设计和更丰富的玩家体验，尤其是在开放世界和沙盒游戏中。影视制作：在影视后期制作中，可用于快速生成特效镜头或预览场景，提高制作效率，同时为创意探索提供更多可能性。# AI工具# AI项目和框架©版权声明本站文章版权归AI工具集所有，未经允许禁止任何形式的转载。上一篇BabyVision - UniPat AI团队推出的多模态理解评测集下一篇司农 - 南京农业大学开源的农业领域大模型相关文章adwrite.ai – AI广告文案写作工具，基于关键词生成多个定制化广告文案AI小集2EduPro – 为留学生打造的AI留学顾问，个性化指导留学过程AI小集4Manga Image Translator – 开源漫画图片文字翻译工具，多语言翻译无缝嵌入原图AI小集7Sheet0 – L4级Data Agent，将任意数据源转为结构化数据表AI小集3LingBot-Depth 2.0 – 蚂蚁灵波科技推出的空间感知模型AI小集2ClawFeed – 开源AI新闻摘要工具，智能提炼结构化摘要AI小集5暂无评论再想想发表评论暂无评论…热门工具Loomy即梦AISekoAiPPT秘塔AI搜索妙呀堆友Agent美图设计室绘蛙AIupdream办公小浣熊秒哒最新收录MetaDigChatArtLexcat造剧立刻成片OJO最新文章秒悟Meoo – 1分钟生成前后端完整的网站，真0门槛开发！42分钟前怎么用 MetaDig 做 AI 短视频，附教程实测案例44分钟前Gemini 3.5 Transcribe – 谷歌推出的最新语音转文本模型4小时前Hy-MT2-1.8B – 腾讯混元推出的端侧翻译大模型4小时前GLM-5.3-Flash – 智谱开源的原生多模态模型，即Ox Alpha10小时前Qwen3.8-Flash – 阿里通义推出的多模态 MoE 模型10小时前Omarchy – 开源的AI原生Linux桌面 ，自然语言生成桌面主题10小时前Play with Putty – 谷歌推出的实时协作 Vibe Coding 工具1天前AZ8 – AI 视频创作工作台，Agent 实时读取素材与状态1天前Tutti · VM – 多人多 Agent 并行协作空间，零延迟并行协作1天前EgoSuite-Open100K – 光轮智能开源的全模态人类行为数据集1天前PixVerse R2 – 爱诗科技推出的实时全模态世界模型1天前Breeze TTS 2 – BreezeBlue 推出的新一代语音合成模型1天前QuickDesk – 开源 AI 原生远程桌面，首款内置 MCP Server1天前AlphaVow – 一站式 AI 批量图像处理工具2天前

PixVerse R2 – 爱诗科技推出的实时全模态世界模型AI工具1天前发布AI小集02PixVerse R2是什么PixVerse R2 是爱诗科技推出的实时全模态世界模型，为PixVerse R1的升级版本。模型支持文字、图像、音频及动作信号等多模态输入，能在运行中持续接收用户控制、实时更新世界状态并连续输出同步音视频。模型通过 Omni Causal AR 架构与实时加速技术，PixVerse R2 在保持长程一致性的同时实现低延迟交互，让视频生成从固定片段进化为可持续演化的动态世界。PixVerse R2的主要功能实时全模态世界生成：支持文本、图像、音频、动作信号等多模态输入，在运行中持续接收用户控制并实时输出同步音视频。长程世界状态保持：通过多时间尺度记忆体系，在长时间运行中维持角色身份、场景设定和关键事件的统一性。动态交互控制：用户可通过 WASD 等动作信号或音频指令，在生成过程中实时操控世界演化方向。自适应时间粒度：根据控制信号的语义边界自动调整音视频生成块大小，兼顾响应速度与表达完整度。错误状态自纠正：通过 Error Bank 机制主动学习并修复历史生成中的偏差，提升长期运行稳定性。PixVerse R2的技术原理全模态因果自回归：将双向时空生成先验转化为单向因果自回归架构，统一处理短视频、长视频与交互轨迹，使画质、音视频表达、长程生成和多模态控制在同一模型中共同扩展。动态分块生成：根据控制信号的时间尺度自适应切分音视频序列，短块提升动作响应精度，长块保留语义完整结构，实现不同任务共享同一因果生成接口。混合教师强制与扩散强制：混合干净历史与带噪历史进行训练，配合因果注意力掩码与相对时间位置编码，缩小训练与推理的分布差距，提升长程稳定性。多时间尺度记忆：由固定锚点记忆、滚动历史记忆和对象键值缓存组成，在固定预算内协同保持世界身份与局部连续性。误差库：将代表性错误历史沉淀为训练样本并回放，使模型主动学习识别和修复偏差，将长期漂移从被动问题转化为主动优化目标。对抗正则去噪分布匹配蒸馏：用全模态因果自回归同时作为学生模型初始化与蒸馏教师，通过条件对齐、分布匹配与对抗正则三信号联合优化，在极少步数下保持控制精度与世界真实感。块稀疏注意力：通过块级相关性路由将注意力稀疏度提升至 90% 以上，集中计算关键依赖，显著降低长上下文开销且效果基本无损。如何使用PixVerse R2目前 PixVerse R2 处于内测阶段，用户需扫码填写体验申请表获取提前体验和 API 接入资格。PixVerse R2的核心优势统一训练范式：采用全模态因果自回归与实时加速层的两阶段统一训练，避免了传统多阶段 Pipeline 的能力折损。全模态实时交互：支持文本、图像、音频、动作信号等全模态输入，能够在生成过程中持续接收用户控制并实时输出同步音视频。长程状态保持：通过多时间尺度记忆体系与误差库机制，在长时间运行中有效保持世界状态统一性并主动纠正漂移。自适应生成粒度：动态分块技术可根据控制信号语义自适应调整生成粒度，兼顾交互响应速度与内容表达完整度。高效无损加速：基于块稀疏注意力与金字塔式超少步蒸馏，在 90% 以上稀疏度下实现低延迟实时生成且效果基本无损。PixVerse R2的同类竞品对比对比维度PixVerse R2Matrix-Game 3.0（昆仑万维）核心定位实时全模态世界模型，面向沉浸式内容与交互体验实时交互世界模型，面向 AI 游戏与虚拟世界构建核心架构Omni Causal AR + 实时加速蒸馏记忆增强的交互式世界模型（Patch 级记忆注入）输入模态文本、图像/视频参考、音频、动作信号（WASD）文本提示、键盘/鼠标动作控制输出内容同步音视频流（Video + Audio）可交互的 3D 游戏世界画面生成性能低延迟流式生成，支持持续交互720p @ 40 FPS（3.0）；单张消费级显卡 20 FPS（3.5）记忆机制多时间尺度记忆（Sink + Rolling + Object KV）+ 误差库纠错Patch 级记忆注入架构，分钟级长程一致性长程一致性通过 Error Bank 主动修复漂移，亮度漂移降低 35.8%分钟级记忆保持，支持数分钟连续探索PixVerse R2的应用场景实时交互式虚拟游戏世界：玩家通过 WASD 等动作信号实时操控第一人称或第三人称视角，模型即时生成对应场景与音画反馈，打造无限延展的开放世界体验。沉浸式影视实时预演：导演在拍摄前通过文本描述和实时动作调整，即时生成分镜动态画面与同步音效，快速验证镜头语言与叙事节奏，大幅降低试错成本。虚拟角色直播与互动娱乐：数字主播或虚拟偶像在直播中实时接收弹幕文本、语音指令或礼物动作信号，动态调整表情、动作与场景，实现真正意义上的实时人机共创内容。实时广告与营销内容生成：品牌方根据用户画像或实时热点，输入产品参考图与简短文案，模型即时生成多版本广告短片并同步配音，支持投放前快速迭代与 A/B 测试。教育培训与仿真模拟：在医学、驾驶或应急演练场景中，学员通过语音或操作设备发出动作指令，模型实时生成对应仿真环境与反馈画面，提供低成本、可重复的沉浸式训练体验。# AI工具# AI项目和框架©版权声明本站文章版权归AI工具集所有，未经允许禁止任何形式的转载。上一篇Breeze TTS 2 - BreezeBlue 推出的新一代语音合成模型下一篇EgoSuite-Open100K - 光轮智能开源的全模态人类行为数据集相关文章CineMaster – 快手推出的文本到视频生成框架，具备3D感知能力AI小集3面试通 – AI面试辅助工具，支持算法、笔试、简历等多种面试模式AI小集3VFusion3D – Meta联合牛津大学推出的AI生成3D模型项目AI小集2HUGWBC – 上海交大联合上海 AI Lab 推出的人形机器人全身控制器AI小集3Vireel – AI视频制作平台，制作易获得关注的短视频AI小集2Asendia AI – AI求职助手，根据个人情况推荐求职策略AI小集2暂无评论再想想发表评论暂无评论…热门工具Loomy即梦AISekoAiPPT秘塔AI搜索妙呀堆友Agent美图设计室绘蛙AIupdream办公小浣熊秒哒最新收录MetaDigChatArtLexcat造剧立刻成片OJO最新文章秒悟Meoo – 1分钟生成前后端完整的网站，真0门槛开发！25分钟前怎么用 MetaDig 做 AI 短视频，附教程实测案例26分钟前Gemini 3.5 Transcribe – 谷歌推出的最新语音转文本模型4小时前Hy-MT2-1.8B – 腾讯混元推出的端侧翻译大模型4小时前GLM-5.3-Flash – 智谱开源的原生多模态模型，即Ox Alpha10小时前Qwen3.8-Flash – 阿里通义推出的多模态 MoE 模型10小时前Omarchy – 开源的AI原生Linux桌面 ，自然语言生成桌面主题10小时前Play with Putty – 谷歌推出的实时协作 Vibe Coding 工具1天前AZ8 – AI 视频创作工作台，Agent 实时读取素材与状态1天前Tutti · VM – 多人多 Agent 并行协作空间，零延迟并行协作1天前EgoSuite-Open100K – 光轮智能开源的全模态人类行为数据集1天前Breeze TTS 2 – BreezeBlue 推出的新一代语音合成模型1天前QuickDesk – 开源 AI 原生远程桌面，首款内置 MCP Server1天前AlphaVow – 一站式 AI 批量图像处理工具2天前Agnes 2.5 Pro Alpha – Agnes AI 开源的多模态推理模型2天前

Gemini 3.5 Transcribe – 谷歌推出的最新语音转文本模型AI工具4小时前发布AI小集02Gemini 3.5 Transcribe是什么Gemini 3.5 Transcribe 是 Google 推出的最新语音转文本模型，支持实时流式和预录音频处理两种模式，前者通过 Live API 实现亚秒级延迟，后者支持说话人归属与词级时间戳。Gemini 3.5 Transcribe 能自动清理填充词与自我纠正、识别 85 种以上语言及实时语言切换、自定义专业词汇，词错误率低至 2.6%，支持调用其他 Gemini 模型完成图像生成等复杂任务。Gemini 3.5 Transcribe的主要功能实时流式转录：通过 Live API 提供亚秒级延迟的连续双向语音转文本服务。预录音频转写：通过 Interactions API 为录音文件提供带说话人归属和词级时间戳的精确转录。智能文本清理：自动去除”嗯””啊”等填充词，修正自我纠正语句，并输出格式化文本。自定义词汇适配：根据用户提供的专业术语和特殊拼写调整转录结果，准确识别邮政编码、订单号等字母数字组合。多语言自动检测：支持 85 种以上语言的自动识别与转录，并能处理实时语言切换和多种口音方言。多说话人区分：在预录音频中准确归属最多三位说话人的发言内容。跨模型功能调用：在转录过程中可调用其他 Gemini 模型执行图像生成、文件分析等复杂任务。屏幕上下文融合：结合设备屏幕内容和对话历史提升特定场景下的转录准确度。Gemini 3.5 Transcribe的技术原理端到端音频理解：模型直接从原始音频波形生成文本，避免中间表示引入的误差累积，同时保留语音中的语调、停顿等副语言信息用于语义消歧。原生多模态融合：基于 Gemini 3.5 统一架构，音频与文本在共享的 Transformer 注意力空间中进行联合编码，使模型能建立声学特征与语义概念的直接映射。上下文感知推理：通过长上下文窗口同时处理语音信号、屏幕截图文本和对话历史，用交叉注意力机制动态加权相关上下文，提升专有名词、文件名称和领域术语的识别精度。增量流式解码：采用自适应分块与投机解码策略，在音频输入持续到达的同时进行局部语义整合，平衡低延迟需求与输出稳定性，支持双向交互中的即时响应与打断处理。如何使用Gemini 3.5 Transcribe开发者接入：通过 Gemini API 在 Google AI Studio 或 Gemini Enterprise Agent Platform 调用模型。实时语音交互：用 Live API 调用gemini-3.5-transcribe-live实现双向流式转录。录音文件处理：用 Interactions API 调用gemini-3.5-transcribe进行带时间戳的离线转写。macOS 桌面端：在 Gemini macOS 应用中直接用语音输入、编辑文本并调用其他模型功能。Android 输入法：在 Gboard 中开启 Rambler 功能，语音输入自动清理并支持语音改稿。Chrome 浏览器：即将支持在任意网页输入框中语音输入与打字。企业部署：通过 Gemini Enterprise Agent Platform 集成至客服与内部工作流。第三方框架：借助 Agora、LiveKit、Vercel 等已接入 Live API 的平台快速搭建语音应用。Gemini 3.5 Transcribe的核心优势转录精度领先：流式场景词错误率低至 4.0%，非流式低至 2.6%，在嘈杂环境中仍能准确捕获字母数字实体。延迟大幅优化：相比前代 Chirp 3，最终转录交付时间缩短 70%，实时流式响应达到亚秒级。智能语义清理：自动去除填充词、修正自我纠正语句，并输出可直接使用的格式化文本。多语言无缝切换：自动识别并转录 85 种以上语言，支持对话中的实时语言切换与多方言口音适配。说话人精准归属：预录音频支持最多三位说话人的发言区分，并附带词级时间戳便于回溯定位。屏幕上下文感知：结合设备屏幕内容与对话历史，显著提升文件名称、专业术语和活跃文档的识别准确度。跨模型任务协同：内置 Function Calling 能力，可在转录过程中调用其他 Gemini 模型完成图像生成、文件分析等复杂操作。Gemini 3.5 Transcribe的项目地址项目官网：https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-5-transcribe/Gemini 3.5 Transcribe的同类竞品对比对比维度Gemini 3.5 TranscribeOpenAI GPT-4o-transcribe定位智能语音转录模型，支持实时流式与预录音频GPT-4o 架构的专用转录模型，API 托管服务词错误率 WER流式 4.0% / 非流式 2.6%约 2.5%（干净音频，行业领先）实时流式延迟亚秒级，原生双向连续流式支持流式，通过 Realtime API 实现智能文本清理自动去除填充词、修正自我纠正、格式化输出不支持，输出原始口语化文本说话人归属原生支持最多 3 人区分 + 词级时间戳不支持原生，需额外调用 diarize 端点Gemini 3.5 Transcribe的应用场景实时会议智能纪要：在多人会议中实时流式转录，自动区分说话人、清理口语填充词，调用 Gemini 模型即时生成会议摘要与待办事项。多语言客服语音助手：通过 Live API 构建亚秒级延迟的语音客服系统，自动识别客户语言并实时切换，结合自定义词汇准确识别订单号、产品型号等关键信息。医疗与法律口述记录：医生或律师通过语音口述病历、庭审记录，模型自动清理自我纠正语句并格式化专业术语，输出可直接归档的标准文档。跨语言实时字幕与翻译：为直播、视频会议提供多语言实时字幕，用 85+ 语言自动检测能力，结合实时语言切换实现无缝跨语言沟通。语音驱动编程与办公自动化：在 Google AI Studio 或 Gemini macOS 应用中，开发者通过语音描述需求即可生成代码，办公人员语音指令调用屏幕上下文完成跨应用文件总结与图像生成。# AI工具# AI项目和框架©版权声明本站文章版权归AI工具集所有，未经允许禁止任何形式的转载。上一篇Hy-MT2-1.8B - 腾讯混元推出的端侧翻译大模型下一篇怎么用 MetaDig 做 AI 短视频，附教程实测案例相关文章Zen Browser – 开源AI网页浏览器，提供深度可定制功能AI小集3Praktika – AI语言学习应用，与逼真的AI数字人互动学习AI小集2bardeen – 零编程构建自动化工作流的AI智能助手平台AI小集3KAEDIM – 3D模型AI生成平台，高质量生成3D模型资产AI小集3AI-Trader – 港大开源的AI自动交易竞赛框架AI小集3豆包视觉理解模型 – 豆包推出视觉理解模型，具备识别和推理能力AI小集31暂无评论再想想发表评论暂无评论…热门工具Loomy即梦AISekoAiPPT秘塔AI搜索妙呀堆友Agent美图设计室绘蛙AIupdream办公小浣熊秒哒最新收录MetaDigChatArtLexcat造剧立刻成片OJO最新文章秒悟Meoo – 1分钟生成前后端完整的网站，真0门槛开发！24分钟前怎么用 MetaDig 做 AI 短视频，附教程实测案例26分钟前Hy-MT2-1.8B – 腾讯混元推出的端侧翻译大模型4小时前GLM-5.3-Flash – 智谱开源的原生多模态模型，即Ox Alpha10小时前Qwen3.8-Flash – 阿里通义推出的多模态 MoE 模型10小时前Omarchy – 开源的AI原生Linux桌面 ，自然语言生成桌面主题10小时前Play with Putty – 谷歌推出的实时协作 Vibe Coding 工具1天前AZ8 – AI 视频创作工作台，Agent 实时读取素材与状态1天前Tutti · VM – 多人多 Agent 并行协作空间，零延迟并行协作1天前EgoSuite-Open100K – 光轮智能开源的全模态人类行为数据集1天前PixVerse R2 – 爱诗科技推出的实时全模态世界模型1天前Breeze TTS 2 – BreezeBlue 推出的新一代语音合成模型1天前QuickDesk – 开源 AI 原生远程桌面，首款内置 MCP Server1天前AlphaVow – 一站式 AI 批量图像处理工具2天前Agnes 2.5 Pro Alpha – Agnes AI 开源的多模态推理模型2天前