- This update brings calendar integration with auto-record scheduling for your meetings! - Chats became multi-document, which lets you ask questions across all your notes at once. - New AI enhancements generate insights and action items from your recordings. - You can now filter transcriptions by speaker, control playback speed, and choose from summarization templates. - Mistral 3 models are now supported. We've improved folder navigation, added echo deduplication for clearer transcriptions, and updated our AI libraries for better performance. Bug fixes include better handling of silent recordings, more reliable audio preprocessing, automatic cleanup of processed audio files, and improved USB audio device detection for call recording.

Local chat with documents! - dive deep into the transcript, chatting with the document, local and secure! AI Transcript Cleanup - Automatically remove filler words and irrelevant segments from your transcripts using AI. Summarization Templates - Create reusable templates for different meeting types with folder and document-level overrides. Enhanced Export - Merge audio tracks into a single file, export as ZIP archives, and see real speaker names in markdown. Redesigned Recording - Large sidebar record button with live timer during recording. Plus: Bug fixes for audio mute state, OpenAI support, and improved tag extraction.

Transcription & Audio - Re-transcribe Segments - Right-click any conversation bubble to re-transcribe just that segment or from that point to the end. Perfect for fixing hallucinations or poor quality transcriptions without re-processing the entire recording. - Click to Seek - Click conversation bubbles to jump directly to that position in the audio playback. Makes it easy to hear exactly what was said. - Delete Conversation Bubbles - Remove individual transcription segments with a right-click context menu option. - Auto-Stop Recording on Sleep - Recordings now automatically stop and finalize when your Mac goes to sleep, preventing data loss. AI & LLM Integration - Full Anthropic (Claude) Support - Complete integration with Anthropic's Claude API as an alternative to OpenAI, Google Gemini, or local models. - Rename Placeholder Participants - AI-generated content now automatically replaces placeholder speaker names (like "Participant-abc123") with actual speaker names when you rename speakers. User Interface - Folder Icons in Sidebar - Visual folder icons now appear in the document sidebar for better organization. - Fluent Audio Model Download Progress - Visual progress indicators when downloading new transcription models. - Explicit Permissions Check - Clearer permission request flow with streamlined UX. - In-App Update Checker - Automatic checking for new app versions with notifications. Organization & Management - Delete Tags and Speakers - Clean up unused tags and speakers directly from the UI. - Speaker Merge Feature - Merge duplicate speaker profiles into a single identity. Bug Fixes - Fixed FluidAudio Confidence Data - Corrected confidence score handling for FluidAudio transcription model. - Preserve User Titles - User-edited document titles are now preserved when re-running summarization. - Rename Untitled Documents - Title field correctly maintained for documents without custom titles. - Performance & Stability improvements

Bug fixes

New Features - Speaker identification - Automatically detect and label different speakers in your transcriptions (requires re-transcribing existing documents) - New Transcription Engine - New transcription engine available in settings for improved accuracy - Folders - Organize your documents into folders for better management - Action Items - View and manage all action items in a centralized page - Import Audio - Import and transcribe audio files directly into Summit AI Notes New Models - GPT-OSS-20b - The most powerful AI model to date, delivering the best results in testing - Granite 4.0 Micro - Two new models from IBM for enhanced transcription capabilities

- Search through transcriptions - Edit transcriptions - Added Large transcription models - Recommend transcription model base on available memory - Delete Audio files - Fixed Custom folders - Fixed Search widget - When using Custom Cloud Provider the model name input stays visible since Anthropic doesn't support the models endpoint - Voice activity detection model will be downloaded automatically if missing