Skip to main content

Changelog

v2.3.7​

Release Date: 2026-09-27

  • Fixed the semi-transparent background rendering of the translation overlay: the overlay texture now uses straight alpha compositing, resolving the issue where the background was multiplied by alpha twice and appeared too dark, achieving opaque text with a semi-transparent background
  • Moved the "Program Startup Settings" card from Advanced Settings to Preferences and updated the navigation labels across all languages
  • Changed "Automatically Enable OCR Image Translation at Startup" to be off by default to reduce resource usage at startup

v2.3.6​

Release Date: 2026-09-26

  • Fixed overlapping text boxes in the PP-OCRv6 model during multilingual recognition

v2.3.5​

Release Date: 2026-09-25

  • Added Text Alignment settings for the translation overlay (Auto / Left / Center / Right)
  • Added Text Color Mode and Background Color Mode (Auto / Custom) for the translation overlay; Auto mode picks colors automatically based on the brightness of the text block background
  • Added local speech synthesis model (Qwen3-TTS) information to the About page
  • Added the ability to start/stop real-time volume monitoring based on the state of the Voice Translation settings page
  • Added a session retry mechanism for online recognition and dual-channel ack matching
  • Fixed misaligned concurrent communication and timeout blocking in the OCR pipeline

v2.3.4​

Release Date: 2026-09-23

  • Added per-application capture for speaker audio: when whole-system loopback capture is unavailable, you can capture only the audio of a specified application

v2.3.3​

Release Date: 2026-09-21

  • Added an OSC message flow direction setting (new messages displayed upward / downward)
  • Added Qwen-Audio-TTS / ASR streaming inference protocol support
  • Fixed an issue where the microphone passthrough TTS configuration became abnormal after upgrading from an older version

v2.3.2​

Release Date: 2026-09-20

  • Added custom endpoint configuration for ASR, supporting real-time / non-real-time recognition modes
  • Added real-time / non-real-time synthesis mode support for TTS
  • Added prompts when the API Key is missing for online recognition and online synthesis, and display of online synthesis errors
  • Unified the recognition state management logic of voice translation

v2.3.1​

Release Date: 2026-09-17

  • Added support for Alibaba Bailian (DashScope) real-time voice translation
  • Added TTS voice presets and voice cloning
  • Added real-time streaming playback for TTS
  • Moved the ASR and TTS settings to the Translation Engine settings page
  • Added TTS switch synchronization for the wrist control panel
  • Added TTS device mapping hints, and a confirmation prompt when switching the TTS inference device to GPU
  • Added subprocess debug log level control
  • Added masking for sensitive information (such as API keys) in logs
  • Fixed voice channel synchronization and playback stuttering issues, and added TTS endpoint protection

v2.3.0​

Release Date: 2026-09-13

  • Upgraded the OCR engine to PP-OCRv6 and merged the previous "Chinese/English/Japanese, Traditional Chinese, Latin scripts" options into a unified Multilingual recognition model
  • Added Qwen3-TTS speech synthesis, which can read out the translation of the microphone channel
  • Added OCR engine preloading to reduce first-screen recognition latency
  • Added A / X key short-press detection and a microphone toggle on the controller
  • Added OCR vertical-merge preset support for the wrist control panel
  • Fixed the OCR translation settings being accidentally overwritten by the wrist panel

v2.2.6​

Release Date: 2026-09-11

  • Added the status icon display feature for the status overlay
  • Changed the wrist control panel / wrist display to be anchored via SteamVR relative transforms
  • Fixed the wheel source directory path in the Python ASR build script

v2.2.5​

Release Date: 2026-09-10

  • Added real-time volume monitoring and volume threshold adjustment
  • Added GPU inference device selection for speech recognition and local translation
  • Added a volume monitoring mode for when the model is not loaded
  • Fixed the voice translation source language to "Auto" and removed the source language selection entry
  • Fixed UI synchronization and prompts when voice translation runs out of memory
  • Fixed the synchronization logic between the voice translation switch and mode
  • Adjusted device enumeration logic and added an integrated-GPU patch

v2.2.4​

Release Date: 2026-09-07

  • Fixed parallax contamination caused by the two eyes capturing frames from different moments
  • Filtered out normal pipe-disconnection error logs from the OCR client

v2.2.3​

Release Date: 2026-09-06

  • Refactored the ASR and OCR subprocesses to reduce memory usage and speed up startup

v2.2.2​

Release Date: 2026-09-06

  • Switched the default translation service to Qwen (Tongyi Qianwen)
  • Optimized the memory check logic to reserve extra memory for the local voice translation model

v2.2.1​

Release Date: 2026-09-02

  • Added system default audio/video device options and improved the microphone / speaker device selection logic
  • Removed the logic that automatically shows the update page on startup
  • Fixed the tutorial panel buttons being obscured by long text

v2.2.0​

Release Date: 2026-09-02

  • Added the Wrist Control Panel feature, letting you overlay a control panel on your wrist with customization for left/right hand, back-of-hand/palm orientation, and display style
  • Fixed the wrist panel not auto-hiding

v2.1.4​

Release Date: 2026-09-01

  • Added microphone and speaker input toggle configuration for voice translation
  • Added SteamVR exit detection and a proper shutdown flow
  • Added a multilingual End-User License Agreement (EULA)

v2.1.3​

Release Date: 2026-08-30

  • Added text block depth mode settings
  • Added a confirmation dialog for switching the speech recognition model
  • Added a confirmation dialog for enabling simulated streaming output
  • Unified the tooltip implementation for dropdown controls

v2.1.2​

Release Date: 2026-08-27

  • Improved per-text-block depth calculation so translation overlays align more precisely with the original text in 3D space
  • Improved the OCR translation overlay display effect
  • Fixed an issue where results could be pinned to the wrist image by mistake while in the "pre-remove" state
  • Unified the tooltip logic for settings controls and aligned all setting items

v2.1.1​

Release Date: 2026-08-25

  • Unified the settings control layout and aligned the right edges of all setting items
  • Unified the slider component and improved the trigger logic after editing
  • Updated the memory threshold required by the local ASR (speech recognition) model

v2.1.0​

Release Date: 2026-08-25

  • Added local voice translation with Hy-MT2 local translation model support, enabling speech recognition and translation even offline
  • Translation models now load on demand and automatically unload when idle, saving memory
  • Adjusted the translation model context limits; added local translation model info on the About page

v2.0.6​

Release Date: 2026-08-24

  • Added localized translation-error tips and improved the error reporting for missing translation API credentials
  • Added font-size adjustment for voice translation subtitles
  • Fixed the voice translation background box adapting incorrectly to resolution changes
  • Fixed overlap caused by a height mismatch in VR translation overlays
  • Added the ability to record the last window state before closing

v2.0.5​

Release Date: 2026-08-23

  • Added a startup window state setting option
  • Optimized theme application timing to avoid flicker when switching
  • Fixed several VR interaction and performance issues

v2.0.4​

Release Date: 2026-08-22

  • Added scroll/animation mode configuration for voice translation results
  • Optimized VRChat Chatbox message processing

v2.0.3​

Release Date: 2026-08-22

  • Fixed an issue where the console failed to hide on Windows 11
  • Optimized the initial size calculation of the Dashboard settings window

v2.0.2​

Release Date: 2026-08-21

  • Added one-click reset for color pickers, and updated the multilingual texts

v2.0.1​

Release Date: 2026-08-21

  • Replaced the ASR engine with a llama.cpp implementation, greatly upgrading the voice translation capabilities
  • Added ASR model selection (Fast 0.6B / High-precision 1.7B) and streaming output
  • Added inference thread configuration for voice translation, improving ASR performance
  • Added single-instance launch protection
  • Fixed screenshot translation being accidentally triggered when the trigger is released
  • Optimized the auto-fit sizing logic of the Dashboard window

v2.0.0​

Release Date: 2026-08-19

This is a major 2.0 release that rewrote the core capabilities around real-time voice translation and restructured the settings UI.

  • Added complete voice translation support, including bilingual translation of in-game speaker audio and your own microphone audio
  • Added sending translation results to VRChat Chatbox via the OSC protocol
  • Added support for multiple LLM translation providers (OpenAI, DeepSeek, Anthropic, Gemini, etc.)
  • Added translation mode switching, remembering the OCR and voice translation modes separately
  • Added interactive onboarding and step-by-step VR in-space tutorials, improving the first-launch flow
  • Added voice translation pause/resume, text alignment, result display mode, and background opacity options
  • Added a status-text display toggle and screenshot-frame opacity adjustment
  • Added controller trigger-position adjustment (front/back, up/down)
  • Optimized the UI and restructured all settings pages

v1.2.1​

Release Date: 2026-07-21

  • Fixed several issues and stability improvements

v1.2.0​

Release Date: 2026-03-19

  • Base version built on OCR image recognition and mainstream translation services
  • Supports in-VR text recognition, translated overlay display, and controller-framed area translation

Older v1.1.x and earlier versions are archived and no longer listed individually.