Smart Transcript Editing Thoth now flags the words it was unsure about and lets you fix them in place. Click a highlight to correct it, or run an AI proofread pass locally, in a verified enclave, or with your own key. Nothing leaves your Mac unless you send it. iCloud Sync Turn it on and your whole library, recordings, transcripts, summaries, and notes, syncs privately across your Macs and iOS devices. Delete on one device and choose whether it goes everywhere or stays where you are. Verifiable Private AI For cloud AI you can now pick a privacy-first provider that runs inside secure hardware enclaves with provable zero data retention. Thoth checks the attestation inside the app before each session. Reminders From Your Meetings Thoth reads the action items out of a transcript and, for the ones you approve, creates real reminders in your Reminders app. Send Recordings Between Devices Move a recording from your Mac to the iPhone app over AirDrop or local WiFi, with no cloud round trip. More Free SRT and VTT subtitle export plus captioned audio video. Anonymization now catches card numbers and IDs with checksum validation and your own custom patterns. Menu bar quick actions, calendar-based meeting detection, and automatic AAC compression for smaller recordings.

Transcript Anonymization Thoth now detects people, organizations, and places, plus phone numbers, emails, and addresses, entirely on your Mac. Before any transcript reaches a cloud AI provider, a preview shows every detected item as a token, so you choose exactly what leaves your device. Switch any transcript between Original and Redacted with one toggle, and right-click a misdetected name to correct it everywhere. A per-recording editor lets you add your own terms, and Apple Intelligence can suggest more. Free for everyone. Privacy should not be behind a paywall. Give Title A new AI action that names a recording from its transcript and renames it in place across your library. Improvements and Fixes The recording detail view now shows which transcription engine produced the transcript, and you can set a preferred engine for re-transcription. AI errors that happen mid-response are surfaced clearly, and a failed run no longer counts against your free monthly AI uses.

Folders and Tags Organize recordings into color-coded folders, up to three levels deep. Assign multiple tags to any recording and filter the list in one click. Apple Speech and Apple Intelligence Two new fully on-device engines. Apple Speech transcribes live sessions and existing recordings across dozens of languages, including mixed-mode (mic and system audio). Apple Intelligence summarizes recordings with no account or API key required. Best suited for shorter recordings. More AI Providers Groq is now available as a cloud AI provider with a generous free tier. Private AI Server connects Thoth to any compatible local server on your network: LM Studio, Ollama, vLLM, and others. Improvements Summary and translation now process the full recording in a single pass. Exported transcripts merge consecutive segments from the same speaker. The language picker now shows only languages supported by the active transcription engine. The live transcript copy button now confirms with a checkmark.

Thoth now lives in your menu bar. Start a recording, pick a mode, switch language, and check recent recordings without leaving your current app. Post-recording actions (re-transcribe, identify speakers, summarize, chat) open in their own window and can be pinned on top. Set global keyboard shortcuts in Settings. When Zoom, Teams, or Slack launches, a nudge appears in the menu bar. One click starts recording in mixed mode, capturing both sides of the conversation automatically. The entire interface is now fully localized in French. The app language follows your system preference automatically. Smaller things: rename speakers directly from the transcript toolbar after diarization, every segment updates throughout. Set a default transcription language in Settings so it is pre-selected on every launch. Toggle background gradients on or off in Settings under Appearance. Imported audio files are no longer deleted from their source location. Bug fixes: notes now save automatically and persist across navigation and app restarts. Cursor and placeholder text in the Notes field are now properly aligned. Renaming a recording is easier: double-click the title or click the pencil icon, and the field gets focus immediately.

Live Transcript now lets you choose the engine. WhisperKit covers all 99 languages. Parakeet TDT V3 hits roughly 190x real-time speed across 25 languages on Apple Silicon. Parakeet EOU goes further: it produces text word by word as you speak, before the sentence ends. English only for EOU, but it is genuinely something to see. AI Chat is now available on every recording. Ask what was decided, who brought up the budget, or get action items broken down by speaker. Answers are grounded in the transcript, not hallucinated. Requires your own API key for a supported cloud AI provider. Share is now one tap from the transcript window, AI result panels, and the notes panel. Mail, Messages, AirDrop, any app on your Mac. Pro users can strip the Thoth footer from everything they share and export. Free tier: the 5-recording cap is gone. Keep as many recordings as you want. AI enhancements also go from 3 to 10 per month. Custom vocabulary now accepts a CSV file and scores multi-word terms together as a phrase, not word by word. Live Transcript now groups incoming chunks into paragraphs with smooth transitions. Much easier to read along. Settings got a full redesign. There is a new Pro section showing your subscription status, quick actions, and social sharing shortcuts all in one place.

Live Transcript now opens as a detachable floating window, so you can keep it visible while switching between apps. Critical bug fix: connecting or disconnecting a Bluetooth device mid-recording silently dropped the rest of the audio. Thoth now reconnects automatically and merges all segments into one clean file, even when your headset switches sample rates. New Parakeet TDT v3 model for re-transcription: faster and more accurate, with support for 25 European languages plus Japanese and Chinese. Available in Settings on Apple Silicon Macs. Thoth now shows an alert if recording is interrupted, so you always know what happened.

• New: tap Report a Bug or Request a Feature in Settings to reach support directly. Includes an optional "how did you find Thoth?" field to help the app grow. • Replaced the Ministral 3B local model with Phi-3.5 Mini Instruct, same speed tier, no login required to download. • Live Transcript: select and copy any snippet directly from the transcript panel during recording.

• Added a Report a Bug button in Settings to reach support directly. • Exports now include a Thoth footer. • Fixed the Reveal in Finder button in the export panel.

French App Store listing added.

Performance & polish: • Fixed high CPU usage when viewing long transcripts and AI summaries. Scrolling is now smooth even on hour-long recordings. • Copy buttons now show a visual confirmation when text is copied to the clipboard. • Summaries are now properly included when exporting with AI enhancements. • RTF exports have been redesigned with professional formatting: justified text, proper headings, paragraph spacing, and clean bullet indentation. • The export panel now shows which AI enhancements are available for each recording.

Free upgrade for all users: • Neural speaker diarization: state-of-the-art models replace the previous audio analysis pipeline. Dramatically better accuracy, especially in overlapping speech. Available to Free and Pro users alike.

• Custom Vocabulary now works at the decoder level: the model actively favors your terms during transcription, not just as a post-processing hint • Local AI summaries are faster: short recordings skip the multi-pass pipeline and are processed in a single shot • Local AI summaries now support French, Spanish, German, Italian, and Portuguese • Settings now shows an estimated memory footprint before you download a local model • Search within any transcript

• Free tier recording limits adjusted: 30 minutes (mic), 15 minutes (system/mixed) • Bug fixes and stability improvements

• System audio and mixed recording modes now available on the free tier (up to 30 minutes) • Microphone recording extended to 60 minutes for free users • Try AI-powered summaries and enhancements. 3 free uses per month • Local AI model can now be downloaded and used by all users • Added "Rate Thoth" option in the Help menu • Bug fixes and stability improvements