OpenVox roadmap.

• Fixed Pause Tags with Punctuation: You can now place pause tags next to periods, commas, and sentence endings without triggering generation errors.
• Crash Fix: Resolved a reported crash during local MLX model compilation on large models audio tasks.

• SillyTavern Support: Added OpenAI-compatible TTS integration for SillyTavern, along with a dedicated step-by-step setup guide.
• AI Dubbing Fix: Fixed an issue where single-speaker projects could use the model's default voice instead of the voice selected by the user when generating the final dubbed output.
• In-App Help & FAQs: Added searchable guides, FAQs, and troubleshooting tips directly from the toolbar for faster help without leaving OpenVox.

• New Sortformer Speaker Identification: Added Sortformer Model as the new default speaker identification model for both AI Transcription and AI Dubbing. Enjoy faster and more accurate multi-speaker separation (upto 4 speakers)
• Fixed an issue where Supertonic 3 preset voices selection bug under Speaker Voices on the Generate & Export step when a specific output language was chosen.
• Improved dubbing segment preprocessing to gracefully handle silence and punctuation-only text without generation errors

This version delivers major improvements to speaker identification, multi-speaker dubbing, and system stability:

• Improved Long-Form Diarization: Stable, high-speed speaker detection for long audio with optimized Apple Neural Engine memory management.
• Advanced Dubbing Controls: Rename speakers across your dub, split dialogue to new speakers, and assign voices with quick-select voice bubbles.
• Improved Speaker Mapping: Refined timeline alignment prevents overlapping dialogue from mislabeling speakers.
• Smarter Voice Sample Detection: Automatically selects the cleanest, highest-energy vocal samples for the speaker from original track.
• Multi-Speaker Script Import: Import SRT and WebVTT scripts with automatic speaker detection in both Dubbing and Conversations.
• UI improvements and bug fixes

Introducing Speaker Diarization, AI Vocal & Background Music Separation, and Multi-Speaker Dubbing, all running 100% locally on your Mac!

• Added MOSS TTS 1.5 for studio-grade 48 kHz local voice synthesis and voice cloning in 31 languages on Mac.
• Added Qwen3 TTS 1.7B for higher fidelity, deeper expressiveness, and improved voice cloning cadence in 10 languages.
• Added an interactive 5-step Help Me Choose wizard and pinned model shortcuts.
• Added inline pause tags ([pause: Xs]) to the OpenAI-compatible Local API.
• Improved memory and threading for cooler, lower-memory long-form generation sessions.
• Includes smoother transitions, hover stability, and improved localized language naming.

• Added Validate Output to locally transcribe generated speech and compare it with the original script.
• Added timestamped review with text-match scores, selective regeneration, and Confirm & Update.
• Improved CJK-language generation and fixed a Supertonic crash issue.
• Voice Clone now accepts MP4 and MOV files and automatically extracts their audio.
• Added precise 0.01x speed adjustments and improved memory handling when switching models.

• AI Dubbing now supports Cloned and Designed Voices directly in the Voice & Output step.
• Improved dubbing timing and added export progress, quick Open and Locate actions, and Export Batch to CSV.
• Improved Batch Mode with one overall progress bar, an Export All completion card, and full-batch cancellation.
• Reduced memory usage during audio merging and fixed output issues affecting 24+ hour audio exports.
• Improved Select & Read player visibility and percentage accuracy on supported devices.

• Added local AI Transcription for audio, video, and direct recording using multiple speech-recognition models.
• Added AI Dubbing with timed scripts, translation workflows, voice selection, and audio/video export.
• Added direct SRT import and translated SRT workflows for dubbing.
• Voice Clone can transcribe reference audio using Apple Speech or downloaded transcription models.
• Voice Changer now accepts video input and exports video with the converted voice track.

• Added drag-and-drop importing across AI Speech, Conversations, Audiobooks, Voice Clone, and AI Voice Changer.
• Improved Smart Numbers & Dates so years such as 1953 are read naturally as “nineteen fifty-three”.
• Improved AI Voice Changer reliability for M4A and MP3 input files.

• Added 200+ new Supertonic voices with support for 30+ languages in Select & Read mode.
• Added avatar images across the Voice Library for easier voice recognition.
• Improved number, date, percentage, currency, and decimal pronunciation with Smart Numbers & Dates, previously called Convert Numbers to Words.
• Fixed previous AI models remaining in memory after switching to Supertonic.
• Improved EPUB imports containing decorative drop caps.
• Improved word replacements containing comma-formatted numbers.
• Added UI fixes and improvements across OpenVox.

• Introducing an all new revamped menu bar with quick access to new actions with resources status.
• Introducing MCP support, making it easier to connect OpenVox with Claude Code, Cursor, and other AI agents.
• Improved Select & Read Page layout and improved handling for larger models and added notch model loading animation.
• Fixed model generation defaults so each model keeps its own parameter settings.
• Fixed Voice Changer respecting the "Auto-play generated audio" setting.
• Fixed audiobook play buttons remaining after generated audio or local data was cleared.

• Added Select & Read to read selected text aloud from any Mac app using a global shortcut.
• Added a compact notch-style player for Select & Read playback.
• Added menu bar mode with quick access to show the app, change voice/speed, and start or stop the local API.
• Added startup options for launching OpenVox in window mode or menu bar mode.
• Added Start OpenVox at Login setting.
• Improved audiobook generation and export progress overlays with clearer chapter and full-book progress.
• Improved audiobook delete options so you can delete the full book or only generated audio.
• Added global word replacement settings under Text Preprocessing, including support for pause tags.
• Fixed multiple bugs in Audiobook workflow.

• Polished UI elements across the AI Speech page and Voice Library.
• Improved AI Audiobook editing with manual chapter creation and drag-and-drop chapter reordering.
• Refined audiobook controls with clearer hover states, better selection styling, and play/pause feedback per chapter.
• Added a larger full-column settings view on the audiobook page for easier model, voice, and generation tuning.

• Fixed a bug where voices did not show on the Audiobook page for Supertonic 3 Model.
• Fixed reported issues in word replacement feature.

• Added Supertonic 3, a blazing-fast multilingual TTS model with 31 languages support that can run up to 2-3x faster than Kokoro.
• Near real-time model loading makes it easier to jump into speech generation instantly.
• Ultra-fast replies are ideal for voice agents, assistants, API workflows, and rapid iteration.
• Great fit for long-form generation where speed matters, including audiobooks and batch voice work.

• Improved Voice Clone trimming with more accurate waveform alignment at all zoom levels.
• Added Local API playback support with play: true, so agents and scripts can ask OpenVox to generate speech and play it directly in the app.
• Improved Local API parameter handling so speed now works consistently across all models.

• Improved Voice Clone with guided trimming for longer audio clips, including waveform preview, zoom, drag handles, and quality checks for clean start/end points.
• Updated Voice Clone sample requirements to 5-12 seconds for more stable results.
• Added a confirmation step to ensure the Reference Script matches the audio clip before saving a cloned voice.
• Improved Voice Changer voice selection so custom voices from all languages appear correctly.
• Added M4A and AIFF source audio selection support in Voice Changer.
• Added model parameter controls to the Local API.

• Fixed a Pocket TTS loading issue that could affect some macOS user accounts.
• Added the ability to create new chapters inside an existing audiobook.
• Chapter title is now editable in the chapter detail editor.

• Added Pocket TTS with support for English, French, German, Italian, Portuguese, and Spanish.
• Delivers fast, low-latency generation for AI agents and Local API workflows using cloned or custom voices.
• Efficiently handles longer generations, including conversations and audiobooks.
• Works with voices from your Voice Library and custom voice recordings.
• Save your preferred model, language, and voice as defaults across AI Speech, Conversations, and Audiobooks.

• Speed controls now support precise 0.05 adjustments.
• Voice Clone now only accepts 5–15 second audio samples to ensure stability of clone voice outputs.
• Voice Design saving is now integrated into the generation results.
• Improved MLX performance tuning for Macs with higher memory.
• Added an interactive guided tour to help explore OpenVox.

• Add precise pauses using tags.
• Save reusable parameter presets for your favourite voice and model combinations.
• Add custom audiobook cover artwork and embed it in exported audiobooks.
• Fixed crashes and improved generation and export reliability.

• Added portable ZIP backup support so you can transfer saved voices between Macs or restore them after reinstalling OpenVox.
• Added an auto-start option for the Local API server, making it easier to use OpenVox with agents, automations, and local workflows.
• Expanded Conversations support from 4 speakers to up to 16 speakers, making it much better for podcasts, scripts, plays, audiobooks, and larger dialogue projects.

• Added M4B export for better audiobook compatibility with chapters and timeline support, and removed the slower MP3 export option.
• Improved reliability and reduced memory usage when merging and exporting large audiobooks, with exports now up to 10x faster.
• Added EPUB cover support, including thumbnail display inside the app.
• Added cover artwork embedding for exported M4A and M4B audiobooks.
• Fixed an issue where batch audiobook generation could show 100% progress before the final chapter had actually finished.

• Added an audiobook export format picker, with M4A selected by default for faster and smaller exports.
• Added real export progress, showing actual bytes processed instead of a generic loading animation.
• Audiobook chapters now use efficient M4A intermediates instead of large temporary WAV files, reducing storage usage by up to 10x.
• Improved cleanup so temporary audiobook files are removed when books or chapters are deleted.
• Added a completion dialog after successful export, with a Show in Finder button to reveal the exported file.
• Polished the export save dialog for a cleaner, more consistent experience.

• Settings are now easier to access from the top title bar, helping you adjust preferences faster and keep your workflow smoother.

• Cloned voices and Voice Design voices now work correctly in Conversations and through the Local API.

• Added OpenVox Local API support for AI agents and external tools.
• Enhanced Voice Templates so gender tags stay in sync with the voice design input.
• Added PDF import in AI Speech single mode for faster text-to-audio workflows.
• Improved word replacements under Pre Processing so they apply more reliably.

• Fixed reversed typing in Batch mode on older Macs.
• Fixed microphone permission detection and recording access for voice cloning.

• Added OmniVoice support for more natural, expressive, and context-aware speech.
• Expanded language coverage to 600+ languages with stronger multilingual performance.
• Added new library voices for 25 more languages using the OmniVoice model.

• Added speed control support to more TTS models beyond Kokoro.
• Improved the AI Audiobook experience across import, chapter handling, and export.
• Added EPUB import for AI AudioBook.
• Large ebook imports now show a proper loading animation to keep long imports responsive.
• Merge and export now include processing feedback for long audiobook exports.
• Individual chapters can now be deleted after ebook import.
• Exported chapter filenames are cleaner and easier to sort.
• Fixed batch paragraph editor text direction behavior for RTL languages on older macOS versions.
• The Daily Power badge is now hidden when Pro is active.

• Added support for custom model storage paths in Settings, including moving existing model files to a new location.
• Batch Mode now supports importing text from .txt, .rtf, and .csv files.
• Added a downloadable CSV template and faster cleanup with the Clear All button.

• Fixed a launch crash affecting users with M1 and M1 Max Macs.
• Model downloads now persist across updates and do not need to be downloaded again after updating.
• Paste now always inserts plain text, avoiding unwanted formatting from browsers and docs.
• Fixed post-processing failures affecting silence removal and normalization.

• Fixed an issue where AI Voice Changer could fail for certain audio clip lengths.
• Improved AI Voice Changer processing for longer audio clips.

• Added Qwen3 TTS support with high-quality reference voice cloning across 10 languages.
• Added Voice Design so you can describe a voice in natural language and generate a brand new reusable voice locally on your Mac.
• Improved output generation times across the app.
• Added a speech recognition helper for Voice Cloning that can detect the spoken reference script from uploaded samples.
• Improved Model Manager download and setup flows.
• Improved model memory management.

• Fixed a Voice Conversion issue where MP3 inputs could fail during processing with a format error.
• Fixed an issue where Manage Models opened the Kokoro model page instead of the main Model Manager page.
• Fixed Resource Manager status so the currently loaded model is shown correctly instead of displaying "no model loaded" when a model is active.