Release 4.4.2

📝 更新日志 / Changelog (v4.4.2)

中文

新增功能

• Gemini Transcribe：Gemini 新增 Transcribe 调用模式
• 识别历史时间轴：历史详情展示录音、识别、润色与文字写入各阶段耗时；弹窗支持按高度自适应

改进优化

• LLM 连通性测试：测试结果分段展示连接、首字、输出耗时与连接复用，并标明是否支持流式
• 识别与键盘响应：异步提交历史、缓存键盘首帧 UI，并更广泛地复用网络连接池，识别延迟更低、上屏更少卡顿、打开更快
• AI 润色文案：界面将「后处理」统一为「润色」
• 打字机效果：关闭后改为润色完成后一次性上屏
• Qwen3-ASR 1.7B 结果清洗：去掉模型标记，避免内部标签写入正文
• 收起键盘中断润色：隐藏键盘时取消进行中的 AI 润色并提交原文，避免下次按麦无响应
• 润色超时与取消：为 AI 润色加上超时上限；取消后不再错误走非流式回退
• [Pro] 热词预加载：启动时预构建音素索引，降低首次识别的后处理延迟

问题修复

• Bridge 模块：优化与 Bridge 模块的通信管理
• 后台通知语言：保活与悬浮等通知将跟随应用语言
• 悬浮球性能：未录音时不再持续刷新悬浮球光环特效

---

English

New Features

• Gemini Transcribe: Gemini adds a Transcribe call mode
• History Timing Timeline: History details show time spent on recording, recognition, polish, and text commit; the dialog adapts to available height

Improvements

• LLM Connectivity Test: Test results show connect, first-token, and output latency, connection reuse, and whether streaming is supported
• Faster Recognition and Keyboard: Async history commit, cached first-frame UI, and broader HTTP connection-pool reuse for lower recognition latency, less commit jank, and faster keyboard open
• AI Polish Wording: UI now uses “polish” instead of “post-processing”
• Typewriter Effect: When off, the full result is applied at once after polish finishes
• Qwen3-ASR 1.7B Cleanup: Strips model markers so they are not committed as text
• Cancel Polish on Keyboard Hide: Hiding the IME cancels in-flight AI polish and commits the raw transcript, so the next mic tap is not ignored
• Polish Timeout and Cancel: AI polish now has a timeout cap; cancelled requests no longer fall back to non-streaming
• [Pro] Hotword Preload: Builds phoneme indexes at startup to reduce first-recognition post-processing delay

Bug Fixes

• Bridge Module: Improved communication management with the Bridge module
• Notification Language: Keep-alive, floating, and similar notifications now follow the app language
• Floating Ball Performance: Aura effects no longer keep refreshing when not recording

Release 4.4.1

📝 更新日志 / Changelog (v4.4.1)

中文

新增功能

• 键盘识别历史面板：键盘扩展按键新增「识别历史」，可在输入法内查看近期记录并插入；支持滑动重新识别或重新后处理
• 保留失败识别记录：识别失败、超时或取消时保留录音，并在历史中展示失败原因，便于重新识别或后处理
• 豆包识别模型选择优化：火山引擎改为直接选择流式/文件 1.0/2.0 模型，不再依赖多个开关组合
• Qwen3-ASR 1.7B Int8：本地 Qwen3-ASR 新增 1.7B Int8 模型
• 输入完成后切换输入法：识别及后处理完成后可自动切换回指定输入法，便于二次编辑
• 常驻通知跳转：开启后台保活后，可选择点击常驻通知打开的页面

改进优化

• 录音重新识别：当选择流式识别模型时，重新识别会自动切换到文件识别引擎或模型，提高重新识别成功率（X-ASR 支持流式回放）
• 供应商设置文案：统一各供应商的识别模型、语言与请求地址等标签
• Pro 介绍：优化 Pro 版介绍页面

问题修复

• 流式预览重复：增加更详细的流式预览过程日志，有相关问题可在导出详细日志后反馈排查
• 滑块拖动生效：滑块仅在松手后写入设置，避免拖动本地模型线程数时反复加载
• 自定义推理参数：修复 LLM 供应商自定义推理参数未正确保存的问题
• 历史录音错绑：取消识别后不再误用上一段录音

---

English

New Features

• Keyboard History Panel: New keyboard extension button to browse and insert recent transcripts; swipe to re-recognize or re-process
• Keep Failed Recognition Records: Failed, timed-out, or cancelled sessions keep audio and appear in history with a reason, so they can be re-recognized or re-processed
• Doubao Model Picker: Volcano Engine now uses a direct model picker for streaming/file 1.0/2.0 instead of stacked toggles
• Qwen3-ASR 1.7B Int8: Added the Qwen3-ASR 1.7B Int8 local model
• Switch IME After Input: After recognition (and post-processing) finishes, optionally switch back to a chosen IME for further editing
• Keep-Alive Notification Tap: When background keep-alive is on, choose which page the persistent notification opens

Improvements

• Recording Re-recognition: When a streaming model is selected, re-recognition automatically switches to a file recognition engine or model to improve success rate (X-ASR supports streaming replay)
• Vendor Setting Labels: Unified recognition model, language, and endpoint labels across vendors
• Pro Intro: Refined the Pro introduction page

Bug Fixes

• Streaming Preview Logs: Added more detailed streaming-preview process logs; export detailed logs if you hit related issues
• Slider Commit: Slider values apply only after release, avoiding repeated local-model reloads while dragging thread count
• Custom Reasoning Params: Fixed LLM vendor custom reasoning parameters not being saved
• History Audio Mix-up: Cancelled recognition no longer reuses the previous audio

Release 4.4.0

📝 更新日志 / Changelog (v4.4.0)

中文

新增功能

• 剪贴板附件同步：剪贴板同步新增图片与文件附件拉取；可监听指定文件夹上传新增文件，支持设置附件大小上限
• 剪贴板文件分享：可将文件通过系统分享进行上传，并增加附件上传/下载通知
• DashScope 新模型：DashScope 新增 Qwen-Audio-3.0-ASR-Flash 流式与非流式模型，移除旧版模型并优化语义标点处理
• [Pro] 热词注入控制：新增热词注入控制的开关，可控制是否将热词注入支持的识别引擎

改进优化

• 使用统计：将使用统计整合为独立的「使用统计」页面，并支持一键生成分享卡片图片
• 界面设置：将界面风格、按键反馈、键盘布局等设置整合为独立的「UI 设置」入口
• IME Bridge 错误处理：增强 IME Bridge 的错误处理与用户反馈提示
• 识别历史UI：优化历史卡片与文本组件的布局与样式
• [Pro] 畅说模式：尝试优化连续对话的会话管理与音频焦点处理，录音更稳定
• [Pro] 热词优化：已包含目标的热词不再重复学习；DashScope 热词管理与同步增强

问题修复

• 优化部分说明文案
• 尝试修复输入法中出现重复流式预览的问题
• 修复搜索空状态内容未居中的布局问题

---

English

New Features

• Clipboard Attachment Sync: Clipboard sync can now pull image and file attachments; it can watch a selected folder to upload new files, with a configurable attachment size limit
• Clipboard File Sharing: Upload files via the system share sheet, with attachment upload/download notifications
• New DashScope Models: Added Qwen-Audio-3.0-ASR-Flash streaming and non-streaming models, removed legacy models, and refined semantic punctuation handling
• [Pro] Hotword Injection Control: New toggle to control whether hotwords are injected into supported recognition engines

Improvements

• Usage Stats: Consolidated usage stats into a dedicated page with one-tap share card generation
• UI Settings: Consolidated interface style, haptic feedback, and keyboard layout into a dedicated UI Settings entry
• IME Bridge Error Handling: Enhanced IME Bridge error handling and user feedback
• Recognition History UI: Refined the layout and styling of history cards and text components
• [Pro] Continuous Talk: Attempted to optimize continuous talk session management and audio focus handling for more stable recording
• [Pro] Hotword Optimization: Targets already present in existing hotwords are no longer learned again; improved DashScope hotword management and sync

Bug Fixes

• Optimized some descriptive copy
• Attempted to fix duplicated streaming previews appearing in the IME
• Fixed the empty search state content not being centered

Release 4.3.0

📝 更新日志 / Changelog (v4.3.0)

中文

新增功能

• 应用快捷方式：长按应用图标可快捷进入识别设置、AI 设置与识别历史
• Fcitx5 / Trime 剪贴板同步：AIDL 联动的 Fcitx5、Trime 修改版支持接入说点啥的剪贴板同步
• [Pro] 自动学习热词：语音输入后修正识别结果并发送可自动学习热词；兼容修改版输入法、IME 桥接与无障碍悬浮球场景

改进优化

• 本地模型下载：支持断点续传与本地缓存，下载更稳、中断后更易恢复
• 下载源测速：下载源测速改用 HTTP 探测
• 设置首页搜索：首页搜索入口与搜索栏样式优化
• 识别历史：优化列表与详情布局
• 本地非流式识别：长音频按静音分段渐进识别，提升本地非流式识别速度，并修复 Qwen ASR 和 FunASR Nano 的本地非流式识别长音频的报错问题
• 配套模块更新：LSPosed IME Bridge、Fcitx5 / Trime 修改版同步更新，配合剪贴板同步与（Pro）热词学习等联动能力

问题修复

• [Pro] AI 助手唤醒词：修复自定义唤醒词被应用语言自动同步强行覆盖的问题
• 收起键盘停录：返回关闭键盘时正确停止录音，避免残留录音状态

---

English

New Features

• App Shortcuts: Long-press the app icon for quick access to ASR settings, AI settings, and recognition history
• Fcitx5 / Trime Clipboard Sync: AIDL-linked Fcitx5 and Trime forks can join SayWhat's clipboard sync
• [Pro] Automatic Hotword Learning: Corrections after voice input can be learned as hotwords, with pending confirmation or auto-add; works with modified IMEs, IME bridge, and accessibility floating ball

Improvements

• Local Model Downloads: Resumable downloads with local caching for more reliable recovery after interruptions

• Download Source Latency: Download source latency checks use HTTP probing

• Settings Home Search: Refined home search entry and search field styling

• Recognition History: Optimizes list/detail layout

• Local Non-Streaming Recognition: Long audio is progressively recognized in silence-based chunks for better local non-streaming recognition speed, and fixes the error of Qwen ASR and FunASR Nano local non-streaming recognition of long audio

• Companion Module Updates: LSPosed IME Bridge and Fcitx5 / Trime forks updated alongside the app for clipboard sync and (Pro) hotword learning

Bug Fixes

• [Pro] AI Assistant Wake Word: Fixed the problem of custom wake word being forcibly overridden by app language automatic synchronization
• Stop Recording on Keyboard Dismiss: Recording stops reliably when the keyboard is dismissed, avoiding a stuck recording state on reopening

Release 4.2.1

📝 更新日志 / Changelog (v4.2.1)

中文

新增功能

• 剪贴板实时同步：SyncClipboard「自动接收」在服务端支持时使用 Realtime 实时同步，否则定时拉取；并可选择保持后台实时连接

改进优化

• Target API 版本：更新 Target API 版本至 36
• 剪贴板同步运行时：IME Bridge 剪贴板同步改为独立后台服务，不再依赖悬浮球运行
• IME Hook 模块更新：优化输入法 Hook 模块设置开关与状态说明；录音区域可用性由桥接模块侧控制；新增多种模块设置，并支持模块设置更改实时生效
• 设置开关说明：更多设置项在开关时展示功能说明

---

English

New Features

• Clipboard Realtime Sync: SyncClipboard Auto Receive uses realtime sync when the server supports it, otherwise periodic pull; optionally keep the realtime connection alive in the background

Improvements

• Target API Version: Updated Target API version to 3
• Clipboard Sync Runtime: IME Bridge clipboard sync now runs as an independent background service and no longer depends on the floating ball
• IME Hook Module Updates: Optimized IME Hook module toggle and status copy; recording-area availability is controlled by the bridge module; added multiple module settings and support for real-time module settings changes
• Settings Switch Explanations: More settings switches show feature explanations when toggled, making on/off effects clearer

Release 4.2.0

📝 更新日志 / Changelog (v4.2.0)

中文

新增功能

• 悬浮球长按录音：悬浮球设置新增长按录音模式，按住开始录音、松手停止
• Cohere ASR：新增 Cohere 语音识别供应商
• 阿拉伯语界面：应用界面新增阿拉伯语语言选项
• 识别历史音频与重跑：识别时可缓存音频；历史详情可查看原文与最终结果，并支持重新识别或重新后处理；其他设置可配置音频保留数量
• IME Bridge 剪贴板同步：剪贴板同步支持通过 IME Bridge，在目标输入法进程存活期间辅助同步

改进优化

• 录音音频焦点：统一录音会话的音频焦点管理，失去焦点时更可靠地停止录音
• 设置页交互：优化对话框退出动画，以及多处 UI 细节

问题修复

• IME Bridge 联动：修复服务关闭时过早取消仍在收尾的会话，避免丢失待交付的识别结果

---

English

New Features

• Floating Ball Hold-to-Record: Added a hold-to-record mode for the floating ball—press and hold to record, release to stop
• Cohere ASR: Added Cohere as a speech recognition vendor
• Arabic UI: Added Arabic as an app language option
• History Audio & Rerun: Recognition sessions can cache audio; history details show the raw transcript and final result, with re-recognize and re-process actions; Other settings can configure audio retention count
• IME Bridge Clipboard Sync: Clipboard sync can use IME Bridge assistance while the target IME process is alive

Improvements

• Recording Audio Focus: Unified audio-focus handling for recording sessions so recording stops more reliably when focus is lost
• Settings Interactions: Smoother dialog exit animations and various UI improvements

Bug Fixes

• IME Bridge Coordination: Fixed premature cancellation of finishing sessions during service shutdown, which could drop pending recognition results

Release 4.1.2

📝 更新日志 / Changelog (v4.1.2)

中文

新增功能

• 输入完成后自动回车：新增开关与扩展按键，识别（或 AI 后处理）完成后自动回车或触发发送，便于聊天应用直接发送
• IME 桥接输入法内录音：可在兼容的第三方键盘中长按录音，并将录音交给说点啥识别（需更新 LSPosed/LSPatch 桥接模块）
• 火山引擎新版鉴权：火山引擎 ASR 支持使用新版 API Key 鉴权方式
• 清除剪贴板记录：其他设置支持一键清除应用内全部剪贴板历史（含置顶项）

改进优化

• 收起后切换输入法：开启「收起后切换」时，切回后的键盘会尝试自动弹出
• 剪贴板预览：剪贴板预览改为有限行数展示，并优化性能
• ASR 设置文案：优化静音自动停止、备用引擎、ITN 等相关说明，表述更清晰

---

English

New Features

• Auto Enter After Input: Added a toggle and extension button to automatically press Enter / trigger the editor Send action after recognition (or after AI post-processing ), making it easier to send in chat apps
• IME Bridge In-IME Recording: Long-press to record inside a compatible third-party keyboard and send audio to BiBi Keyboard for recognition (requires an updated LSPosed/LSPatch bridge module)
• Volcengine New Authentication: Volcengine ASR now supports the new API Key authentication method
• Clear Clipboard History: Other settings can clear all in-app clipboard history at once (including pinned items)

Improvements

• Switch IME on Hide: When “switch IME when hidden” is enabled, the keyboard will try to reappear after switching to the target IME
• Clipboard Preview: Clipboard preview is now a limited number of lines display with more stable text normalization and truncation
• ASR Settings Copy: Clarified descriptions for silence auto-stop, backup engine, ITN, and related settings

Release 4.1.1

📝 更新日志 / Changelog (v4.1.1)

中文

信息更新

• Pro 版调价：Pro 版价格从 4.49 美元调整为 5.49 美元

---

English

Information Update

• Pro Price Adjustment: The price of Pro has been adjusted from $4.49 to $5.49

Release 4.1.0

📝 更新日志 / Changelog (v4.1.0)

中文

新增功能

• StepAudio 端点配置：StepAudio 新增请求端点预设，支持按量付费、Coding Plan 与自定义端点
• MiMo / StepAudio 分端点 API Key：MiMo 与 StepAudio 支持按端点预设分别保存 API Key
• 本地备用模式：备用引擎设置新增本地备用驻留模式，可在按需加载与保持常驻之间选择
• [Pro] 悬浮球畅说模式：Pro 版悬浮球支持畅说模式，可在悬浮球场景下持续监听与分段识别
• [Pro] IME 桥接输入上下文：Pro 版启用 IME 桥接的悬浮球录音也可使用输入框上下文辅助 AI 后处理

改进优化

• 悬浮球录音视觉反馈：录音时根据音量显示光晕与峰值波纹，处理中视觉同步增强
• 备用引擎策略重构：重构备用 ASR 仲裁与超时逻辑，本地备用支持懒加载串行策略，并新增启动/识别状态提示，体验更流畅
• VAD 判停优化：输入音量大小归一化处理，改善各引擎 VAD 表现，判停相关设置文案更清晰
• 一键设置无障碍检测：优化无障碍服务需求判断，更准确反映悬浮球、音量键与 IME 桥接的实际配置

问题修复

• 调试日志清理：修复关于页调试日志导出时清理功能不生效的问题
• 备用引擎容错：空结果不再被视作调用失败，提升备用引擎切换逻辑的容错性

---

English

New Features

• StepAudio Endpoint Configuration: Added endpoint presets for StepAudio, supporting pay-as-you-go, Coding Plan, and custom endpoints
• MiMo / StepAudio Per-Endpoint API Keys: MiMo and StepAudio now support saving API keys separately for each endpoint preset
• Local Backup Residency Mode: Added local backup residency settings to choose between on-demand loading and resident mode
• [Pro] Floating Ball Continuous Talk: Pro floating ball now supports continuous talk mode for persistent listening and segmented recognition
• [Pro] IME Bridge Input Context: Pro floating ball recording with IME bridge enabled can now use input box context for AI post-processing

Improvements

• Floating Ball Recording Visual Feedback: Shows halo and peak ripple effects based on audio levels during recording; enhanced processing visuals
• Backup ASR Strategy Refactor: Refactored backup ASR arbitration and timeout logic with lazy serial local backup support, plus loading/recognition status indicators for a smoother experience
• VAD Auto-Stop Optimization: Input volume normalization for better VAD performance across engines; clearer auto-stop setting labels
• One-Click Setup Accessibility Detection: Improved accessibility requirement detection to better reflect floating ball, volume key, and IME bridge settings

Bug Fixes

• Debug Log Cleanup: Fixed debug log cleanup not working when exporting logs from the About page
• Backup Engine Tolerance: Empty results are no longer considered as call failures, improving the fault tolerance of backup engine switching logic

Release 4.0.3

📝 更新日志 / Changelog (v4.0.3)

中文

新增功能

• 可见时持续录音：键盘或悬浮球可见期间在本机持续录音，按麦后更快开始识别
• 录音超时停止：新增超时停止录音自动停止模式，并支持配置最长录音时长
• Soniox 识别模式：Soniox 流式识别新增低延迟到高精度的灵敏度选项

改进优化

• OpenAI 自定义端点兼容性：自定义服务不再生效非流式压缩
• 悬浮键盘拖动把手：根据窗口底部间隙自动将拖动把手切换到顶部或底部
• 麦克风按钮样式：去掉麦克风按钮固定方形限制，支持更灵活的自定义布局比例
• 自定义键盘布局网格同步：调整任一面板网格大小时同步更新所有面板，并自动过滤越界按钮
• 优化备份设置页文字样式

问题修复

• X-ASR 尾部音频：修复 X-ASR 流式识别停止时可能丢失尾部音频的问题
• [Pro] Pro 备份修复：修复 Pro 版配置备份丢失部分内容的问题

---

English

New Features

• Continuous Recording While Visible: Records audio locally while the keyboard or floating ball is visible for faster recognition startup
• Recording Timeout Stop: Added timeout-based auto-stop recording mode with configurable maximum recording duration
• Soniox Recognition Mode: Added low-latency to high-accuracy sensitivity options for Soniox streaming recognition

Improvements

• OpenAI Custom Endpoint Compatibility: Non-streaming compression no longer applies to custom endpoints
• Floating Keyboard Drag Handle: Automatically moves the drag handle to the top or bottom based on bottom window gap
• Microphone Button Styling: Removed the fixed square microphone button constraint for more flexible custom layout proportions
• Custom Keyboard Layout Grid Sync: Syncs grid size across all panels when any panel is resized, automatically filtering out-of-bounds buttons
• Improved backup settings page text styling

Bug Fixes

• X-ASR Tail Audio: Fixed potential loss of tail audio when stopping X-ASR streaming recognition
• [Pro] Pro Backup Fix: Fixed partial content loss in Pro configuration backup