UI 与交互全面升级

• iPhone 使用 iOS 26 原生 Liquid Glass 标签栏与页面过渡，切换标签时会保留各页面的导航状态
• 技能页改为更清晰的已安装与未安装列表；点按技能进入完整详情页，集中管理安装、快捷指令配置和自定义技能编辑
• 重新整理首页、账号、AI 服务和个人词库中的信息层级；缺少必要配置时，会直接说明原因并提供下一步入口

新增个人专属风格

• 可从自己的听写表达与明确修改中学习说话风格，生成专属于你的措辞和表达习惯
• 个人风格可应用于剪贴板回复，在不改变原意的前提下调整用词与节奏，让回复更像你自己
• 用于学习的文字语料仅在你主动点击生成时发送给所选 AI 服务，生成结果需要确认后才会保存和启用

新增智能剪贴板技能

• 端侧模型与本地规则会对剪贴板内容进行基础语义识别，并从完整技能目录中智能推荐最多五个相关操作
• 可识别任务、邀约、问句、情绪和语言，推荐回复、翻译、总结、澄清、共情、商务回复或整理清单等技能
• 复制单个电话号码时可直接拨号或新建联系人；复制网页链接时可直接打开，或提取公开网页内容进行总结
• 识别和推荐过程在设备端完成；只有主动运行需要 AI 的技能时，相关内容才会发送给所选服务

新增自动推荐个人热词

• 根据键盘中反复输入的新词自动生成个人热词推荐，让常用名称、术语和表达更容易再次输入
• 推荐词必须由你明确确认后才会加入个人词库；本地学习不会记录完整击键内容

UI and Interaction, Rebuilt

• iPhone now uses the native iOS 26 Liquid Glass tab bar and transitions while preserving each tab's navigation state
• Skills are organized into clearer Installed and Not Installed lists; tap one to manage installation, Shortcut setup, or custom-skill editing on a dedicated detail page
• Home, Account, AI Service, and Personal Dictionary now have a clearer hierarchy, with direct explanations and next steps when setup is incomplete

New Personal Style

• Learn a personal speaking style from your dictation and explicit edits, capturing the wording and expression patterns that make your voice distinct
• Apply that style to clipboard replies so wording and rhythm sound more like you without changing the original meaning
• Learning text is sent to the selected AI service only when you explicitly tap Generate, and the result must be reviewed before it is saved and enabled

New Smart Clipboard Skills

• On-device models and local rules perform lightweight semantic recognition, then recommend up to five relevant actions from the complete skill catalog
• Tasks, invitations, questions, sentiment, and language can surface skills such as Reply, Translate, Summarize, Clarify, Empathy, Business Reply, or Organize List
• A single copied phone number can open calling or a prefilled new-contact form; a copied web link can open directly or be summarized from bounded public webpage content
• Recognition and recommendation stay on device. Content is sent to the selected service only after you run a skill that requires AI

New Automatic Personal-Term Suggestions

• Repeated new terms from keyboard input become suggestions, making frequently used names, terminology, and expressions easier to type again
• A suggestion joins Personal Dictionary only after your explicit confirmation; local learning does not record complete keystrokes

新功能

全新 App 界面

• 首页改为信息总览：使用统计、听写历史和个性词库集中显示为资料卡，点按后使用标准返回按钮进入完整列表
• iPhone 底部导航重做为 首页 / 技能 / 风格 / 设置四栏 Liquid Glass Dock，绿色胶囊清楚标示当前页面
• 技能与润色风格统一为卡片式管理：点按选择、右上角编辑、右下角显示启用状态；技能支持像主屏幕图标一样长按拖动排序
• 设置入口重新整理，AI Agent、剪贴板、语音与文字输入等页面更容易找到，当前设置摘要直接显示在导航行右侧
• iPad 改为侧栏 + 内容区布局；键盘横竖屏均充分利用可用宽度，并加入系统地球键、撤销、重做、复制和剪切等操作

中英文输入全面升级

• 中英文输入均可在本机学习用户的输入习惯、候选频率和选择偏好；设置中可单独清除打字习惯，不会删除个性词库
• 拼音优化混合简拼排序，并支持 zh、ch、sh 两键简拼
• 英文键盘新增类 QuickType 候选栏，同时显示原词、纠错和补全；约 4 万词离线词表会结合系统词库、通讯录名称与文本替换提供建议
• 英文邻键纠错更贴合真实键位，并减少对专名、短词和全大写词的误改；拒绝纠错后会学习保留原词
• 支持叠指连打、英文双空格句号，以及根据输入框显示前往、搜索、发送、完成、下一项等回车键文案

更稳定、更可控

• 听写、AI 回答、编辑结果和剪贴板粘贴可通过统一的撤销操作回滚
• 修复切换键盘崩溃、中文输入初始化后无法恢复、通用剪贴板导致键盘卡住，以及 AI 会话结束后麦克风不可用等问题
• 全部内置与自定义润色风格都会保留问句原意，不再把用户的问题改写成回答
• 设备端语音识别仍为默认选项；AI 使用用户自己配置的服务商与 API Key，不再提供内置 DeepSeek 回退
• 长按麦克风现用于向 AI 提问；编辑上次输入改用右下角编辑按钮。旧的长按剪贴板语音指令已由剪贴板历史和快捷技能取代
• 已安装旧中文名称“待办”或“日程”快捷指令的用户，需要从技能页重新安装 OSGExtractTodos 和 OSGExtractEvents
• 已移除空白区域滑动光标及其设置开关

New Features

A New App Interface

• Home becomes a true overview: usage statistics, dictation history, and Personal Dictionary appear together as preview cards that open full lists with standard back navigation
• The iPhone dock is rebuilt around four Liquid Glass destinations — Home / Skills / Styles / Settings — with a green capsule marking the current page
• Skills and polish styles share one card-based interaction: tap to select, edit from the top-right, and see enabled status at the bottom-right; skills reorder with a Home Screen-style long-press drag
• Settings navigation is reorganized so AI Agent, Clipboard, Speech, and Text Input are easier to find, with current summaries shown directly beside each row
• iPad adopts a sidebar + detail layout. The keyboard fills the available portrait and landscape width and adds the system globe key, undo, redo, copy, and cut

Chinese and English Typing, Rebuilt

• Chinese and English typing learn candidate frequency and selection preferences locally. Typing habits can be cleared without deleting Personal Dictionary entries
• Pinyin improves mixed abbreviations and supports two-key zh, ch, and sh abbreviations
• The English keyboard adds a QuickType-style bar for the typed word, correction, and completion; an offline list of about 40,000 words works with the system lexicon, contact names, and text replacements
• Neighbor-key correction follows the physical keyboard while avoiding unwanted changes to names, short words, and all-caps text; rejecting a correction teaches the original
• Overlapping key presses, the English double-space period, and contextual Go / Search / Send / Done / Next Return labels make typing feel closer to the system keyboard

More Stable, More Controllable

• One Undo action now rolls back dictation, AI answers, edits, and clipboard pastes
• Fixes include keyboard-switch crashes, Chinese input recovery after setup, Universal Clipboard freezes, and a microphone that could remain unavailable after AI sessions
• Every built-in and custom polish style preserves questions instead of rewriting them as answers
• On-device speech recognition remains the default. AI uses your configured provider and API key; the built-in DeepSeek fallback has been removed
• Holding the microphone now asks AI, while Edit Last Input moves to the bottom-right Edit button. Clipboard History and quick skills replace the old hold-to-command clipboard flow
• If you installed the older Chinese-named Tasks or Events Shortcuts, reinstall OSGExtractTodos and OSGExtractEvents from the Skills page
• Blank-area cursor sliding and its Settings toggle have been removed

新功能

润色模式整体升级

• 大幅升级趣味润色（含直男癌等），风格更鲜明，同时减少事实失真
• 润色强度简化为轻度 / 重度：轻度优先保真，重度释放五种趣味风格的完整创意
• 新增同音 / 近音纠错，语音识别后的错字修复更准
• 长文本润色与翻译分段更清晰，列表结构更易读，翻译后保留段落格式

问题修复

• 修复润色代答问句：口述问句（如「你能听到我说话吗？」）时不再被改写成一句回答，全部风格与强度都会保留问句本身
• 修复自定义润色风格编辑偶发显示空白模板

稳定性与可靠性

• 重构蓝牙与音频路由处理：切换蓝牙设备（耳机）后，录音恢复更可靠
• 修复录音启动 / 取消竞态，减少识别完成后卡在润色阶段的问题
• Apple 本地识别改用降噪接口：同房间旁人说话时，更倾向听清你的近讲声音
• 为识别与润色增加明确超时与原文兜底，网络不稳时不再无限等待
• 加强主 App 与键盘的结果确认与恢复，降低结果丢失或无法上屏的概率
• 修复火山 ASR 连接问题；修复云端 ASR 配置验证误报
• 重构键盘启动流程、降低内存峰值，减少闪退与被系统终止
• 输入引擎按需加载，语音与文字切换更快、更省内存
• 未使用麦克风时，不主动拉起主 App 占用内存
• 移除冗余多步引导；优化会话未完成时的体验
• 自定义词 / 本地识别相关能力打包更完整

New Features

Polish upgrades

• Stronger fun polish styles (including Straight-Talk Rescue / 直男癌), with clearer voice and less factual drift
• Intensity simplified to Light / Heavy: Light prioritizes fidelity; Heavy unlocks the full creativity of the five fun styles
• New homophone / near-homophone correction after speech recognition
• Clearer segmentation for long polish and translation; lists read better; translations keep paragraph structure

Bug Fixes

• Fixed polish answering a question: dictating a question (e.g. “Can you hear me?”) is no longer rewritten into an answer — the question is preserved across every style and intensity
• Fixed the custom polish style editor occasionally opening a blank template

Stability & Reliability

• Rebuilt Bluetooth / audio-route handling so recording recovers more reliably after switching headsets
• Fixed start / cancel races that could leave the UI stuck in polish after recognition finished
• Apple on-device recognition uses voice-processing / noise reduction so your near-talk mic is preferred over other speakers in the room
• Explicit timeouts and original-text fallback for recognition and polish — no endless waits on flaky networks
• Stronger host ↔ keyboard result confirmation and recovery to reduce lost or uninsertable text
• Fixed Volcengine ASR connection issues and false cloud ASR config-validation failures
• Rebuilt keyboard launch path with lower peak memory to reduce crashes and jetsams
• Input engines load on demand — faster Voice ↔ Typing switches and lower memory use
• Does not wake the host app just to hold memory when the mic isn’t used
• Removed redundant multi-step prompts; smoother experience when a session isn’t finished
• More complete packaging for custom words and on-device recognition assets

新功能

• 新增「润色风格包」：内置轻度清理、清晰结构、正式表达、日常聊天，并支持 iCloud 同步
• 新增「趣味润色风格」：直男癌拯救器、装逼指南、大厂黑话、帝吧大神、小红书集美五种润色风格
• 新增自定义润色风格功能
• 新增「画中画」保活，并作为默认风格，支持在设置中切换回灵动岛；关掉画中画即可结束会话
• 云端识别支持真流式（百炼 / 火山 / OpenAI），说话时文字会实时上屏
• 新增历史记录可按天删除（只清当天）
• 设置里对流式 ASR 供应商显示「流式识别」标签；OpenAI 云端识别默认走流式

体验改进

• 设置页重新设计，结构更清晰
• 灵动岛保活默认改为 5 分钟（更省隐私与电量），支持多种时长设置
• 更新 iPhone 历史、词库 Tab 图标，与 Mac / iPad 统一
• 润色风格卡片与设置摘要更简洁，少装饰、少冗余状态字
• 设置、风格等相关页面统一卡片留白与圆角
• 更新权限与隐私说明

问题修复

• 完成新手引导后不再卡在最后一页
• 删除重装后会重新出现欢迎引导
• 松手后尾音更不容易丢字；识别过短时会整句再识别一次兜底
• 按下录音时麦克风预热阶段的开头语音更不易被丢掉
• 末段识别为空时，不再把前面已经识别好的字抹掉
• 修复趣味润色在短句时的失真问题

New Features

• Polish style packs: Light Cleanup, Clear Structure, Formal, and Casual Chat, with iCloud sync
• Fun polish styles: Straight-Talk Rescue, Flex Guide, Big-Tech Buzzwords, Tieba Legend, and Xiaohongshu Vibes
• Custom polish styles
• Picture-in-Picture keep-alive as the default, with Dynamic Island as an option in Settings; ending PiP ends the session
• True streaming cloud ASR (Bailian / Volcengine / OpenAI) with live on-screen text while you speak
• Delete history by day (clear only that day)
• Settings shows a “Streaming ASR” badge for streaming providers; OpenAI cloud ASR defaults to streaming

Improvements

• Redesigned Settings with a clearer structure
• Dynamic Island keep-alive defaults to 5 minutes (more privacy- and battery-friendly), with multiple duration options
• Updated iPhone History and Dictionary tab icons to match Mac / iPad
• Simpler polish-style cards and Settings summaries — less decoration and fewer status words
• Consistent card padding and corner radius across Settings and style pages
• Updated permission and privacy copy

Fixes

• Onboarding no longer sticks on the last page after completion
• Welcome onboarding shows again after delete-and-reinstall
• Less end-of-utterance truncation on release; very short captures retry the whole phrase
• Speech at the start of mic warm-up is less likely to be dropped
• An empty final segment no longer wipes earlier recognized text
• Fixed fun-polish distortion on short sentences