Repository navigation
0.2.6
- Fixed wrist dialogue panels staying visible when empty or idle, set the default auto-hide timeout to 15 seconds while preserving saved settings, and renamed the setting to “Auto-hide” with “Keep visible” shown for zero.
- Fixed cloud speech recognition stalling during reconnection, added buffering for up to 30 seconds of audio, and improved handling of queued audio on stop and individual segment failures.
- Limited VRCX speech recognition hints to member names to prevent unwanted world names in transcripts, while preserving spoken world names and applying hint setting changes to active capture sessions.
- Fixed dictionary imports blocking subtitle history writes, and kept the previous dictionary available if a replacement fails or is cancelled.
- Corrected subtitle history storage accounting so dictionary and learning data no longer trigger premature history deletion.
- Added automatic recovery after VR overlay worker failures and reduced per-frame processing overhead.
- 修复手腕对话面板在无内容或空闲超时后仍保持显示的问题,默认自动隐藏时间改为 15 秒并保留已有设置,同时将设置项改名为“自动隐藏”,零值显示为“保持显示”。
- 修复云端语音识别重连时卡住的问题,支持缓存最多 30 秒音频,并改进停止识别时的剩余音频处理和单段识别失败后的恢复。
- VRCX 语音识别提示仅使用成员名称,避免世界名称被误写入字幕,同时保留实际说出的世界名称,并让提示设置的修改应用于正在运行的采集会话。
- 修复导入词典时阻塞字幕历史写入的问题,替换失败或取消时保留原词典。
- 修正字幕历史的存储占用统计,避免词典和学习数据导致历史记录被提前清理。
- 增加 VR 叠加层工作线程异常后的自动恢复,并减少每帧处理开销。
Full Changelog: 0.2.5...0.2.6