Fixed: a downloaded Cantonese model could disappear from the language list after you left and reopened the app. content updated: The Apple model now recommends iPhone 15 Pro or newer. Clearer credit rules: Plus credits (180 a month) do not roll over, top-up credits do not expire, and recording uses Plus credits first.

New • AI Summary : pulls the key points and to-dos out of your transcripts. • AI Proofread : finds and marks misheard words. Simpler pricing • "Remaining Time" is now "Cloud Credits". 1 credit = 1 minute. Your balance stays the same. • Plus now costs HK$58 a month and includes 180 cloud credits every month. (Existing monthly subscription members will be issued compensation uniformly. If you are omitted, please report the issue through the app interface -> My -> Suggestion Feedback -> Report Problem) • Top-ups are now 60 and 180 credits. • Local models stay free and unlimited. Improvements • Polished parts of the interface. The model picker now lets you compare supported languages and credits per minute at a glance. • Switching models and languages is smoother, and your recording keeps going. Fixes • Fixed some display errors in the interface. • Fixed some buttons that did not show their correct state. • Fixed Plus still showing after a subscription expired.

Bug fixes • Monthly subscription: starting August 15, a backend fault let purchases go through without unlocking the paid features. This is fixed. Affected accounts have been repaired, with 1 hour of cloud model time added as compensation. If we missed your account, please tell us through Feedback. • Restore Subscription now shows progress and tells you the result. One invalid receipt no longer hides an active subscription behind it. • Silence auto-stop now works with the Apple on-device model. • The Feedback list now says when it failed to load, instead of showing an empty list. • Corrected wording in several places. The Plus page now states that speaker identification works on the cloud model only. New • Transcripts are kept on this device. Save now stores the transcript instead of opening a share sheet, and a recording is saved automatically when it ends. Open More ▸ Transcripts to read one in full or export it as a text file. Saving is a Plus feature; exporting a transcript you already have is not. • Every model states its limits before you pick it. The on-device model loses accuracy in a noisy room, and takes a few seconds to load each time you start. The cloud model needs a network connection. The waiting screen now says which of the two waits you are in. • A guided tour for new users. Six pages on the one thing the app never explained: it can caption the room around you, or the audio playing on your phone. Replay it any time from Settings ▸ How to use. • The History button on the control bar is now called Usage, so it is no longer confused with Transcripts.

Update UI Fix bugs Update speech-to-text quality

correct the issue of accurately receiving sound but no subtitles update UI

Auto Picture-in-Picture: captions now pop into a floating window automatically when you switch to another app — no extra taps needed Stability and UI update

new ui style Improve fluency

新功能: - 增加"使用時長紀錄"功能 - 更新部分UI - 語音轉文字現在支持背景聲音獲取（不包括通話）、畫中畫功能 語音轉文字Model: - 由於效能並未符合預期，移除Apple Acoustic 引擎模型 Bug fix： - "清除"按鈕現在可以被點擊使用

根據用戶反饋： 現在可選擇使用「Apple 神經網路引擎」或「聲學引擎」，運算完全在裝置端進行，無需連網。 新增本地模型選擇，並易於管理各語言包。 進階功能（如說話者識別）會根據用戶當前選擇的語音模型自動調整顯示，避免混淆。 優化部分繁體中文表達意思。

修正語音轉文字功能長時間會停止服務的BUG 新增“用戶反饋”功能