Live Transcription (Beta) is here! See your words appear in real time as you record — fully on-device and offline, or streamed through ElevenLabs. And it's more than a transcript: • Conversation Copilot — Live AI cues as you talk: nudges, knowledge lookups, name context, and fact checks. • Live AI Chat — Ask questions about your conversation while you're still recording. • Context & Notes — Pre-load meeting notes, key facts, and people so every suggestion stays relevant. Also in this update: • Redesigned AI Providers — Pick a provider for each feature (Transcription, Summary, and Chat) in one place. No more separate Cloud/Offline/Hybrid modes. • Performance improvements and bug fixes. Note: On-device Live Transcription is still in Beta and may be unstable on some devices — switch to ElevenLabs if you run into issues.

► ElevenLabs Live Transcription — Real-time streaming transcription while recording using ElevenLabs, with account credit usage display in Settings > AI Mode ► Apple Intelligence — Use Apple's on-device Foundation Model for both Summary and Chat. Select 'Apple' from the provider dropdown in Settings > AI Mode. Private, fast, and requires no API key (iOS 26+) ► Separate Summary & Chat Providers — Summary and Chat now have independent provider selection. Choose different models for each task from the new dropdown menus in AI Mode settings ► Groq default model updated — Switched from deprecated Llama 3.3 70B to OpenAI GPT-OSS 20B for faster and more cost-effective text generation

► Keyterm prompting for ElevenLabs — Improve transcription accuracy for names, brands, and jargon by adding comma-separated terms in Settings > Transcribe Prefs > Keyterm Prompt ► Fixed a rare crash that could occur when processing completes while the History list reloads ► Re-summarize and Re-transcribe options now available directly from the transcript menu ► Switch between plain and segmented transcript views from the menu

► ElevenLabs Scribe v2 — New transcription model option with improved accuracy, better multilingual support, and enhanced handling of accented speech ► Improved speaker diarization reliability for long multi-speaker recordings ► Fireworks AI default model updated to Kimi K2 for improved text generation quality ► Added option to disable rating prompts in Settings > About

► Improved transcript handling for summary and chat accuracy ► Removed Fireworks AI from transcribing provider (deprecated by provider) ► Added advanced in-app search and replace function for individual transcriptions ► Fixed an issue where incoming phone calls would interrupt and stop an active recording ► Minor UI improvements and bug fixes

► Fireworks AI — New cloud provider for transcription and text generation ► Cloud AI Model Preferences — Browse, select, or enter custom models for each cloud provider directly from Settings ► Text Input — Paste or share text directly into the app for summarization and AI chat ► Provider labels now shown on transcriptions and summaries so you always know which model was used ► Minor UI improvements and bug fixes

► Now Free to Download — Try the app with a Demo (Lite) mode before upgrading ► Existing Paid Users — Automatically upgraded to Pro seamlessly with all features unlocked (no action needed) ► Demo Mode — Record and transcribe up to 2 minutes with all features available ► Redesigned UI — Fresh, modern interface with streamlined navigation throughout the app ► New App Icon — A sleek new look that reflects the app's modernized design ► Clear Chat History — Reset AI chat conversations to start fresh with a clean context ► Video Import — Upload videos directly from your photo library and extract audio for transcription ► Batch Selection — Select multiple transcriptions in History for bulk organization or deletion ► Deepseek Provider — Added Deepseek as a new cloud option for AI chat and summarization ► Overall performance improvements and bug fixes

► Enhanced AI Chat - Chat now uses both transcript AND summary for more comprehensive answers when using cloud providers ► Chat Without Summary - Chat is now available even without generating a summary first - just need a transcript ► Local AI Optimization - Smart context limits for local models to fit within ~2K token windows ► Re-Generate Transcription - Easily re-transcribe any recording with different settings or providers for better accuracy ► Video to Audio Conversion - Share video files (MP4, MOV, etc.) and the app automatically extracts the audio for transcription ► Smart Recording Limits - Recording duration is now automatically limited based on your device's memory to prevent crashes (30 min to 4 hours depending on device) ► Audio Files in Files App - Recordings are now saved to your device's Files app (On My iPhone > AI Transcribe) for easy backup and access ► Report Bug/Crash - New button in Settings to easily report bugs and crashes with automatic device diagnostics and memory usage info ► Minor UI improvements and bug fixes

Fixed broken links for LlamaCpp models.

• Fixed minor UI bug - Model management buttons now respond to a single tap instead of requiring a long press. • Added timestamp export from History - Share transcripts with timestamps directly from the history list menu. • Optimized sharing performance - Text and transcript sharing is now instant with no delays. • New LlamaCpp Model - Added Clinician Note 2.0a model, a medical text generation model fine-tuned from Qwen3. Designed for clinical note synthesis and summarization. Not intended for clinical use without validation. • Minor UI and performance improvements.

► Added Whisper prompting support - Improve accuracy by adding a short prompt with any names, terms, or context the audio includes. ► UI fixes and performance improvements.

► Fixed playback speed selector - Speed can now be changed instantly during playback with a single tap ► Fixed audio duration display - Duration now shows correctly immediately when opening transcriptions

► Fixed an issue where iCloud Sync caused UI freezes for some users ► In-app audio recordings are now automatically saved to History after stopping, improving reliability ► Updated Google Gemini models to the latest versions for better AI chat and summarization qualit

• iCloud Sync - Keep your transcriptions, AI conversations, summaries, and audio files available on every device - work anywhere, pick up where you left off • Smart Organization - Automatically categorize recordings by type (Meeting, Interview, etc.), mark favorites, and create custom folders to keep everything organized • Titles are now editable - Tap any transcription title to customize it • Start recording instantly from anywhere - Home Screen widgets, Control Center, or quick actions with a long-press on the app icon • Minor UI and performance improvements

• ElevenLabs Transcription - Added ElevenLabs as a new cloud transcription provider (free tier offers 10,000 credits monthly) • Speaker Identification - ElevenLabs uniquely identifies different speakers in conversations with color-coded labels (diarization) • Fixed an issue where 'Transcribe Only' mode still required AI summarization models even though they weren't being used • Optimized local AI chat responses by reducing context size and improving token efficiency • Added ability to copy and share transcripts timestamps • AI chat transcripts from conversation history can now also be copied and shared • Fixed an issue where audio playback could not be controlled when app was in the background • Improved handling of contents sharing and copying • IReplaced Claude Sonet 3.5 with Claude Sonnet 4 for Anthropic BYOK • Minor UI and performance improvements for iOS 26

• Fixed an issue with file upload from other locations • Fixed an issue where MP3 audio files cannot be played back under transcript history • Minor UI and performance improvement

- LlamaCpp Summarization - Optimized for more accurate and human-like summaries • New LlamaCpp Models - Specialized models for text summarization — a must-try • Background Recording - Continue recording when switching apps or locking phone • Pause & Resume - Take breaks without losing recording progress • Edit Segment Transcripts - Fix mistakes or refine specific parts of your transcripts • Timestamp Export - Share transcripts and summaries with time markers included

• Segmented transcripts with timeline – Easily navigate through new transcriptions with a structured timeline. • Transcript search – Quickly find keywords across all your saved transcripts. • Performance boost – Faster processing with WhisperCpp and LlamaCpp models. • Improved reliability – Fixed issues related to network loss during cloud transcription. • Fresh new look – Enjoy a complete UI overhaul with a compact menu and a redesigned app icon.

• Optimized Whisper model processing and removed iOS memory allocation limits, unlocking the full potential of devices with 8 GB of RAM for transcribing with Whisper (Thanks to Arthur) • Switched from MLX Swift to Llama.cpp for more accurate offline summaries, with several new LlamaCpp models available. • Added the ability to manually select downloaded models (Whisper/Llama) instead of automatic selection. • New option to transcribe audio without generating a summary. • Introduced Google AI as an additional text generation provider (free API key). Performance improvements and bug fixes for a smoother experience.

## New Features • Added Language Configuration Panel on Home page for easier language selection • Added Whisper Large V3 Turbo model for faster transcription • Added 5 new text generation models including ultra-light options for all devices ## Improvements • Optimized language processing capability for better accuracy • Models with high memory requirements now show warning buttons instead of being disabled • You can now download any model after accepting a memory warning (Note: Using these models WILL definitely crash the app. Our app automatically calculates your device's RAM against model's requirement when you see the purple buttons) • Improved model selection to automatically use the best available model ## Bug Fixes • Fixed audio playback not playing through speaker • Fixed issue where app showed one model selected but used a different one • Fixed model priority to correctly select higher quality models first ## Notes • Purple download buttons indicate models that may use significant memory • The app will warn you before downloading large models but allows you to proceed if desired (Using them will DEFINITELY crash the app, unless you somehow add more RAM to your device) • New ultra-light models available for devices with limited memory

## Language Preference & Translation - **Manual Language Selection**: Choose from 25+ languages for more accurate transcription instead of relying on auto-detection - **Translation Mode**: New toggle to choose between: - Keep Original Language: Transcribe and summarize in the audio's native language - Translate to English: Automatically translate everything to English during transcription - **Smart AI Integration**: Titles and chat questions now respect your language preference - **Improved Accuracy**: Manual language selection provides better results than auto-detection ## Share Target Support - **Easy Audio Sharing**: Share audio files directly from other apps to Transcribe - **Seamless Integration**: Select Transcribe from the share sheet in any app that supports audio sharing - **Quick Processing**: Instantly start transcribing shared audio files without manual importing ## Other Bug Fixes ## Tips - For best results with non-English audio, manually select the language in Settings - Use "Translate to English" mode when you want everything (transcript, summary, and chat) in English - Check out the new Language Settings tutorial in the app for detailed guidance