Skip to content

v2.0.3

Choose a tag to compare

View all tags
@github-actions github-actions released this 09 Jun 09:11
· 16 commits to master since this release
v2.0.3
280f053

v2.0.3 更新内容（Updates）

配音能力 / Dubbing

  • 新增高质量配音流水线，支持字幕解析、文本清洗、配音规划、TTS 生成、音频拼接和视频合成。
    Added a high-quality dubbing pipeline with subtitle parsing, text cleanup, dubbing planning, TTS generation, audio assembly, and video muxing.

  • 支持短字幕自动合并、口播时长估算和配音时间线对齐，提升配音自然度和同步效果。
    Supports automatic short-subtitle merging, spoken-duration estimation, and dubbing timeline alignment for more natural and synchronized output.

  • 支持在文本过长时通过 LLM 改写为更适合口播的表达。
    Supports LLM-based rewriting when text is too long, making dubbing scripts more suitable for natural speech.

  • 新增 [dubbing] 配置项，可调整字幕合并、语速范围、文本改写等配音参数。
    Added [dubbing] configuration for subtitle merging, speed limits, text rewriting, and other dubbing parameters.

字幕样式 / Subtitle Styles

  • 新增字幕样式自定义能力，支持通过 JSON 文件配置横屏/竖屏字幕样式。
    Added customizable subtitle styles, allowing horizontal and vertical subtitle layouts to be configured via JSON files.

  • 支持分别配置主字幕和副字幕的字体、字号、颜色、描边、阴影、边距、对齐方式等。
    Supports separate styling for primary and secondary subtitles, including font, size, color, outline, shadow, margins, and alignment.

  • CLI 新增 --subtitle-style-file 参数，可在字幕生成和视频渲染时指定样式文件。
    Added the --subtitle-style-file CLI option for specifying custom styles during subtitle generation and video rendering.

  • 新增默认字幕样式和示例样式文件，方便用户参考修改。
    Added default and example subtitle style files for easier customization.

封面生成 / Cover Generation

  • CLI cover 命令正式接通，可通过提示词生成视频封面图。
    The CLI cover command is now fully wired and can generate video cover images from prompts.

  • 支持自定义封面尺寸，并保存最终提示词和生成图片到任务目录。
    Supports custom cover sizes and saves both the final prompt and generated image to the task directory.

稳定性优化 / Stability Improvements

  • 优化 SRT 字幕解析，支持 UTF-8 BOM，并增强非法时间段校验。
    Improved SRT parsing with UTF-8 BOM support and stricter invalid timestamp validation.

  • 优化静音字幕识别和文本清洗，避免误删括号内容、连字符文本等有效口播内容。
    Improved silence detection and text cleanup to avoid removing valid spoken content such as parenthesized text or hyphenated expressions.

  • 增强配音规划、TTS 生成、音频拼接、字幕样式加载和渲染的稳定性。
    Improved robustness across dubbing planning, TTS generation, audio assembly, subtitle style loading, and rendering.

  • 优化 CLI dry-run 行为和帮助文案，使参数校验流程更清晰。
    Refined CLI dry-run behavior and help text to make validation-only flows clearer.

Assets 19
Loading