优化（Optimization）

• 更好的时间戳匹配算法
  • Better timestamp matching algorithm
• 更新部分界面文本和布局，修改导航项及按钮标签
  • Update some UI text and layout, modify navigation items and button labels

修复 (Fix)

• 修复原本在 Windows 下 edge-tts 的下载错误
  • Fix the download error of edge-tts on Windows

Full Changelog: v1.3.1...v1.4.0

新增（New Features）

• 配音新增EdgeTTS支持
• Added EdgeTTS support for dubbing.
• 桌面版添加了黑夜模式，server修改ui和桌面版保持一致，server版也添加了配置页
• The desktop version now includes a dark mode. The server UI has been updated to match the desktop version. The server version also now includes a configuration page.
• 添加FasterWhisper 的GPU加速选项（50系必须开启此选项才能正常使用FasterWhisper ）
• Added a GPU acceleration option for FasterWhisper (50-series GPUs must enable this option to use FasterWhisper properly).

优化（Optimization）

• 优化翻译上下文，翻译更准确
• Optimized translation context for more accurate translations.

新增（New Features）

• 全新切分&翻译逻辑，任务失败率以及对LLM能力要求都大幅降低
• Completely revamped segmentation & translation logic, significantly reducing task failure rates and lowering the requirements for LLM capabilities
• 对亚洲语言翻译结果中的符号进行处理，观看更美观
• Improved handling of symbols in translations for Asian languages, resulting in a more polished viewing experience
• 二次切分过长字幕，实现影视级单行字幕
• Secondary segmentation for overly long subtitles, achieving cinematic-grade single-line subtitles

修复（Fixes）

• 部分情况下Windows文件路径导致任务失败
• Fixed task failures caused by Windows file paths in certain scenarios

此版本配置文件有变化（新增app.max_sentence_length），请对照最新的config/config-example.toml.
This version includes changes to the configuration file (added app.max_sentence_length). Please refer to the latest config/config-example.toml.

新增（New Features）

• LLM增加JSON开关，如果模型支持json format可以打开，提高输出内容准确度
  Added JSON toggle for LLM. If the model supports JSON format, enabling this can improve output accuracy.

修复（Fixes）

*字幕过长的问题（二次分割）

• Fixed issue with overly long subtitles (secondary segmentation)
  *使用本地视频时合成视频失败
• Fixed failure in video synthesis when using local videos
  *某些特殊的文件名导致ffmpeg报错
• Fixed FFmpeg errors caused by certain special filenames

优化（Optimization）

• 优化分段处理音频分割逻辑
• Improved audio segmentation logic in chunk processing
• 时间戳生成算法
• Enhanced timestamp generation algorithm
• 生成更多任务中间结构的本地文件便于排查问题
• Generated more intermediate task structure files locally for easier issue troubleshooting

此版本配置文件有变化（LLM配置增加json项），可见桌面版配置页面或config-example.toml。
Note: This version includes configuration changes (added JSON option in LLM config). Refer to the desktop version configuration page or config-example.toml.

原项目名KrillinAI更改为KlicStudio，方便记忆与传播，谢谢支持~
The original project name KrillinAI has been changed to KlicStudio for better memorability and dissemination. Thank you for your support~

Hotfix

• 修复新版task id在输入源为youtube链接时出错的问题
• Fixed an issue where the new task ID would fail when the input source was a YouTube link
• 修复新版阿里云语音识别句子处理错误的问题
• Fixed a sentence processing error in the new version of Alibaba Cloud speech recognition

v1.2.1版本用户请直接切换到此版本，v1.2.1版本的具体更新内容请见上一个release。
Users of v1.2.1 can directly switch to this version. For the specific updates in v1.2.1, please refer to the previous release.

新增（New Features）

• 语音合成支持OpenAI TTS
• Added OpenAI TTS support for voice synthesis
• 桌面端任务失败后，增加失败信息提示并停止状态查询
• Desktop app now displays failure messages and stops status polling when a task fails
• 桌面端配置页移除保存按钮，切换tab和关闭软件时会自动保存
• Removed the "Save" button in desktop app settings; changes are now auto-saved when switching tabs or closing the app
• 桌面端配置变化后，在启动任务前会自动重启服务端确保配置都生效
• Desktop app now automatically restarts the backend service before task execution to ensure configuration changes take effect
• 勾选合成视频后，生成的语音现在会合入最终的视频
• Generated voice tracks are now merged into the final video when "Synthesize Video" is enabled

优化（Optimization）

• 重新组织配置文件结构，更便于理解
• Reorganized configuration file structure for better readability
• 任务id及任务文件夹名现在会基于上传的文件生成，增加可识别性
• Task IDs and folder names are now generated based on uploaded files for better identification
• 重写阿里云语音合成，提高任务稳定性
• Rewrote Alibaba Cloud voice synthesis module for improved task stability
• 优化纯音乐音频片段的处理
• Enhanced handling of pure music audio segments
• 调整默认的并发配置值
• Adjusted default concurrency configuration values
• 优化配置项文案
• Improved configuration item descriptions

修复（Fixes）

• 音频克隆源文件上传失败
• Fixed audio clone source file upload failures
• 桌面端代理配置不生效
• Fixed desktop app proxy configuration not taking effect
• 没有开启双语字幕时配音失败
• Fixed voice synthesis failures when bilingual subtitles were disabled

此版本配置文件结构有较大调整，请参照最新的config-example.toml，更新前最好备份一下原来的配置文件。
Note: This version includes significant changes to the configuration file structure. Please refer to the latest config-example.toml. It's recommended to back up your existing configuration file before updating.

新增（New Features）

• web&桌面版支持选择多个本地文件，依次处理
• Multi-file Support: Web and desktop versions now support selecting multiple local files for sequential processing.
• 支持分别配置转录和翻译的并发数，提高任务处理效率
• Concurrent Configuration: Supports setting separate concurrency numbers for transcription and translation to improve task processing efficiency.
• 支持配置转录和翻译的最大重试次数
• Retry Configuration: Supports configuring the maximum number of retries for transcription and translation.

修复（Fixes）

• 某些情况下勾选合成字幕嵌入视频后，出现ytdlp相关报错
• Subtitle Embedding Issue: Fixed ytdlp-related errors that occurred in some cases when embedding synthesized subtitles into videos.

此版本配置文件有变化，web版请参考最新的config-example.toml，桌面版请关注配置页面转录和翻译相关配置项（有默认值）。
The configuration file has been updated in this version. For the web version, please refer to the latest config-example.toml. For the desktop version, please pay attention to the transcription and translation-related settings on the configuration page (default values are provided).

修复（Fixes）

• 字幕嵌入横竖屏选项在桌面端不生效
• Subtitle orientation (landscape/portrait) option not working on desktop
• 配音连接阿里云服务超时
• Voiceover service connection timeout with Alibaba Cloud
• 某些情况下输出的文本文件是空
• Empty output text files in some cases

其它（Others）

• 日志打印移除配置内容，防止密钥不小心泄露
• Removed configuration details from logs to prevent accidental key leakage
• 大模型输出内容和原文不匹配的时候不再中断流程，改为输出warn日志
• No longer interrupts process when LLM output doesn't match source - now outputs warn logs instead

新增

• 文本转录本地模型支持whispercpp（当前只支持windows系统，后续再补充）
• 源语言支持马来语
• 目标语言扩充至101种

修复

• 桌面端黑暗主题看不见语言选择项

其它

• 桌面端源语言默认选项调整为英文
• 更新了视频下载器安装源，之前版本过旧了（视频下载失败的话可以尝试把bin下面的ytdlp文件删掉然后重启软件安装新版）
• Docker部署问题修复

• 修复 桌面端遗漏server配置项
• Fixed the missing server configuration item on the desktop version.
• 修复 桌面端下载语音文件失败
• Fixed the issue where the desktop version failed to download audio files.
• 修复 桌面端切换左侧tab时任务进度丢失
• Fixed the issue where task progress was lost when switching tabs on the left side of the desktop version.
• 修复 桌面端可能出现api路径错误的问题
• Fixed the potential issue of incorrect API path on the desktop version.