1. Improved offline upload reliability for long audio files. 2. Added server-side audio verification to ensure uploads are fully available before completion. 3. Fixed an issue where shared files (AirDrop) could create duplicate upload tasks. 4. Better background retry behavior for unstable networks. 5. Improved multilingual status messaging during upload and transcription.

1, Offline transcript viewing fixed Open a history record without network and still see the real-time transcription text instead of “No transcript content.” 2, Smoother navigation during recording Switching to Settings and back while recording no longer causes the app to hang or freeze. 3, General stability improvements Better background-state handling and reduced UI thread load for a more responsive experience.

1, Improved long-audio upload reliability with automatic background retry for unstable networks. 2, Added an upload task queue so files are processed in the background without blocking the UI. 3, Added per-file upload progress so you can track each file in real time. 4, Completed uploads are now removed from the queue automatically for a cleaner experience. 5, Failed upload tasks now retry intelligently and resume automatically when reopening the upload screen with a stable connection. 6, Settings now save locally first and sync to server later, so changes are not lost when connectivity is poor. 7, Improved overall app stability and performance during extended transcription sessions.

1, Improved stability for long recordings — Fixed an issue where the app could freeze or become unresponsive after 10+ minutes of continuous transcription. 2, Better online engine reliability — The online speech engine now automatically reconnects if the connection drops mid-session, with no need to restart recording. 3, Fixed session timeout — Resolved an issue where transcription silently stopped after approximately 10 minutes due to a server-side connection limit. 4, Reduced lag during extended sessions — Significantly improved performance after 5+ minutes of use; the app no longer accumulates internal overhead that caused delayed responses. 5, More accurate usage tracking — Fixed a billing calculation error that could cause recordings to stop earlier than expected based on your subscription plan. 6, Engine switching reliability — Fixed a bug where switching between transcription engines could cause them to interfere with each other.

VIP users can now use cloud speech recognition for real-time transcription. Includes bug fixes and improved usage tracking accuracy.

• Word-level audio sync: tap transcript text to jump to that moment in the recording • Speaker rename: rename all occurrences across transcript with one tap • AI chat: ask questions about your recording directly in the app • Improved sync with Nvoice web app

Bug Fixes 1, Offline transcription now works for all languages (English, Japanese, Korean, etc.) 2, Fixed real-time transcription for some languages 3, Fixed native events being silently dropped

1, Added Real-time Interpreter — press and hold to speak, get instant translation and voice playback in the target language. 2, Face-to-face mode: rotate the top speaker panel 180° so both parties can read their own translation panel comfortably. Toggle with the rotate button in the header. 3, Interpreter sessions are automatically saved to history with duration and audio when you leave the page. 4, Updated Help & FAQ with Real-time Interpreter usage guide.

1, Added four live AI trigger modes: Keywords, Select, AI, and Enter. 2, Improved real-time Chinese punctuation for finalized Sherpa transcription. 3, Enhanced sentence segmentation and readability for long transcript chunks. 4, Reduced punctuation duplication and improved live-session stability. 5, Added instant settings sync: changes to AI mode, keywords, and summary options now take effect immediately during transcription and meetings. 6, Improved offline punctuation accuracy for Chinese and English.

Improved AI Chapter Generation Fixed an issue where AI-generated chapters could be regenerated repeatedly even when chapters already existed. Chapters are now reliably preserved across all sessions. Referral Program Updates Updated referral descriptions to clearly explain the bonus minutes reward: your friend receives 200 bonus minutes upon sign-up, and you earn 200 bonus minutes once your friend has cumulatively used 10+ minutes. Stability & Bug Fixes Resolved various edge cases that could cause incorrect behavior when loading transcript details General performance improvements and UI polish

Smarter Hold-to-Ask 1, Fixed an issue where ambient noise or brief accidental taps could trigger the Hold-to-Ask feature unintentionally. The button now requires a minimum hold duration and checks that real speech was captured before sending to AI. 2, Improved noise detection supports both English and Chinese — short fragments or background sounds are automatically filtered out. Better Audio Recognition 1, Fixed a bug where the noise suppression mode (intended only for Hold-to-Ask) could persist into normal recording sessions, causing audio from external sources to go unrecognized. 2, Normal transcription now receives clean, unfiltered audio at all times. Punctuation Improvements 1, English transcriptions now include automatic punctuation (periods and question marks) powered by iOS speech recognition. 2, Chinese transcriptions now receive punctuation at natural sentence boundaries — questions end with "？" and statements end with "。". Bug Fixes & Stability Various improvements to audio session handling between Hold-to-Ask and normal recording modes.

AI Chat Improvements 1, AI responses now render with full Markdown formatting — bullet points, bold text, headings, and code blocks display correctly in both the live recording chat and history detail chat 2, Added Nvoice assistant avatar in the real-time chat for a cleaner, more conversational look 3, AI responses in history records no longer show "Searching the web…" or similar status lines — only the final answer is displayed Stability & Performance 1, Fixed a crash caused by multiple concurrent speech recognition sessions starting simultaneously — the app now guarantees only one recognition session runs at a time, preventing CPU overload 2, Improved Voice restart logic: back-to-back speech end and error events no longer trigger duplicate restarts; any pending restart is cleanly cancelled and replaced

1, Streaming AI Responses: AI chat now delivers answers in real time with token-by-token streaming, so you see results instantly instead of waiting for the full response. 2, Rich Markdown Formatting: AI responses are now rendered with full Markdown support including headers, bullet lists, bold/italic text, and code blocks. 3, Hold-to-Ask: Press and hold the microphone button inside any AI chat to ask a question hands-free. Your spoken question is transcribed and sent to the AI assistant automatically.

Bug Fixes & Performance 1, Fixed an issue where transcription text appeared duplicated after stopping a recording 2, Significantly reduced app size by removing unnecessary bundled resources (~200MB smaller) 3, Fixed a crash that could occur when starting voice recognition due to audio session conflicts 4, Improved language detection to ensure the correct speech engine is used for each language Improvements 1, Faster and more reliable real-time transcription for all supported languages 2, Optimized text rendering during continuous recording sessions

Audio Playback Cache: Audio files are now cached locally. Re-opening a transcript loads audio instantly without re-downloading from the server. AI Content Reliability: Fixed an issue where AI-generated fields (summary, outline, keywords, title) were not properly saved or were repeatedly regenerated each time the transcript detail page was opened. Data Integrity Improvements: Fixed multiple issues where history records could lose data during sync, ensuring all fields are correctly preserved across screens. Offline Transcription AI Processing: Improved detection of completed offline transcriptions that are missing AI content, ensuring summary, outline, and keywords are properly generated. Smarter Content Detection: The app now correctly identifies placeholder values (e.g., "Updating...") and raw audio filenames as invalid titles, automatically replacing them with AI-generated alternatives. Stability & Performance: Various bug fixes to prevent redundant network requests and eliminate data race conditions between screens.

Performance & Stability • Migrated Chinese speech recognition engine from WebView-based WASM to Native iOS framework, significantly reducing memory usage and eliminating out-of-memory crashes on older devices • Optimized audio processing pipeline for faster and more reliable real-time transcription Background Recording • Recording now continues seamlessly when the screen locks or the app enters the background • Added automatic transcript recovery — your work is auto-saved in case of unexpected interruptions Bug Fixes • Fixed an issue where the speech recognition engine was not activated on the second and subsequent recordings • Fixed duplicate text appearing in the transcript when stopping a recording • Improved reliability of mid-session engine switching for Chinese transcription • General stability improvements for both solo recording and online meeting modes

iPad Support 1. Full native iPad support with optimized landscape split-view layout 2. Side-by-side transcription and AI chat panels on iPad 3. All orientations supported Faster Transcription 1. Significantly reduced startup delay for offline Chinese transcription (Sherpa ONNX) 2. Optimized recognizer lifecycle — no more re-initialization between sessions 3. Faster endpoint detection for quicker text output Improved Hold-to-Ask AI 1. Fixed text capture accuracy — questions sent to AI now precisely match what you spoke 2. Extended noise suppression stabilization period for cleaner audio input 3. System-level noise cancellation activates during Hold-to-Ask Bug Fixes & Stability 1. Fixed offline transcription resource loading issue 2. Fixed network module initialization crash 3. Improved background recording reliability

1. Hold to Ask AI: Press and hold the new floating button during recording to speak your question. Release to send it to AI with full transcript context for instant answers. 2. Faster Transcription Start: Speech recognition engine now pre-initializes in the background, reducing startup delay when you hit Record. 3. AI Chat Reliability: Fixed "Service interruption" errors when using the AI assistant. Conversations now load and save correctly across devices. 4. VIP Usage Display: Fixed an issue where VIP members' usage limits displayed incorrectly.

1. Fixed AI chat — conversations now work reliably 2. Chat history is now properly saved and synced across sessions 3. Added Locate feature: quickly jump to and highlight selected text in transcript 4. Added text selection popup with Copy, Explain, and Find Related actions 5. Improved transcript segment display and audio seek accuracy

1. Word-Level Audio Sync: Tap any word in the transcript to jump to that exact moment in the audio. Words highlight in real time as audio plays. 2. Speaker-Segmented Transcript: Clear speaker labels and timestamps for each segment, with speaker rename and inline text editing. 3. Interactive Progress Bar: Tap anywhere to navigate through the recording. 4. Bug fixes and performance improvements.

Fixed an issue where changes to Scenario Mode in Settings would not save correctly — switching presets (e.g. from "Default" to "Meeting") now persists as expected after saving

Improved History Sync Transcription records now appear in your History immediately after saving, with AI-generated titles and keywords updating in real time — no manual refresh needed. Enhanced AI Language Accuracy AI-generated summaries, chapters, and chat responses now more reliably match the language of your original transcript. Usage Tracking Improvements Fixed an issue where usage data could be calculated incorrectly under certain network conditions. Usage calculations for real-time recordings and meetings are now more accurate. Stability & Performance Improved data synchronization reliability across all transcription modes. General bug fixes and performance enhancements.