• New: Agent mode — an on-device AI assistant with tools, MCP connections, and saved chats • New: Custom Pipeline (beta) — build your own transcription, translation, and voice pipeline • Interpreter mode: new conversation mode (private playback for you, loudspeaker for your partner) and a politeness/register control • Added Moonshine on-device speech recognition and rebuilt native text-to-speech with voice cloning • Regenerate transcript summaries anytime, with your own custom instructions • TV Mode now has adjustable caption size and a more compact header • Faster AI translation for saved transcripts • Fixed live audio routing bugs and improved language detection in Interpreter mode • Apple Watch app improvements and bug fixes

- Ability to select transcription and translation model in conversation mode - Improved error handling, bug fixes on Apple Watch companion app

This update brings major improvements to live translation and voice features. • Added new live translation workflows with expanded realtime transcription support (gpt-realtime and qwen-livetranslate) • Added Face-to-Face mode on iPhone for two-way translated conversations on a single device (qwen-livetranslate) • Added a keyboard dictation extension for faster text input and caption workflows • Improved voice cloning with separate voices for each language for more natural translated playback • Added state consistency when relaunching the app for last used languages selected

• New transcript dubbing playback controls for translated transcripts • Improved live translation workflows and realtime transcription • Added shadowing, JLPT vocabulary, Japanese annotation, and grammar tools • Expanded app audio and conversation features with continued stability improvements • Polished layouts, playback behavior, and overall reliability

Fix proximity pairing stuck: remove NISession, guard double-invite

• New Conversation mode for two-person live translation, with both chat and face-to-face layout, and app to app (p2p) pairing. • Improved UI/UX and settings

-bug fix for translation not working in certain scenarios.

This latest release includes the following changes: 1. Improves the usability and performance of the conversation mode Auto Detect turns feature in conversation mode. 2. Summarization on whole conversation 3. Simplified Conversation UI

New in this version: • Caption audio from other apps on iPhone with Screen Broadcast • New full-screen TV mode for easier reading at a distance • Improved live translation with Apple and AI model options • Better conversation flow, speaker handling, and overall stability

This update adds a major new Conversation mode, deeper Japanese learning tools, Apple Watch support, and better transcript playback. • New Conversation mode for two-person live translation, with both chat and face-to-face layouts. • Automatic spoken playback of translated captions when headphones are connected. • Synced translation playback for saved transcripts, so translated lines can be spoken in time with recorded audio. • New Grammar Lesson feature for transcript lines, with follow-up questions and support for OpenAI, Anthropic, Gemini, or local OpenAI-compatible models. • Expanded Japanese learning support with furigana, romaji, part-of-speech coloring, and improved reading accuracy. • Added Apple Watch support for live caption viewing, remote session controls, and watch microphone streaming to iPhone. • Improved transcript handling with audio recording, saved conversation support, transcript import options, and cleaner translation grouping. • Better live caption stability and translation quality, including punctuation handling, context-aware translation, and fewer duplicate or regressive lines.

Fixes bug where language selector label was stacked vertically on some phone screens.