Skip to content
Back to skills

Voice Subtitle Sync

ASecurity

Prepare or repair voice and subtitles using authorized recordings or configured TTS, measured audio timing, transcript checks and SRT validation.

  • 4 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 22, 2026
designpython

Security analysis

A100/100

Pro scans all 6 files and shows the line behind each finding

Scanned October 22, 2026

npx -y skills add VisualxIntelligence/motion-tools-and-skills --skill voice-subtitle-sync --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Voice Subtitle Sync?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Voice Subtitle Sync
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/visualxintelligence-voice-subtitle-sync/badge)](https://www.skillsdirectory.com/skills/visualxintelligence-voice-subtitle-sync)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: voice-subtitle-sync
description: Prepare or repair voice and subtitles using authorized recordings or configured TTS, measured audio timing, transcript checks and SRT validation.
---

[English instructions](SKILL.en.md) · 简体中文

For English-language tasks, read [SKILL.en.md](SKILL.en.md) and its English reference templates before proceeding. Both editions use the same tools and authorization boundaries.


# 配音与字幕同步
先判断是已有原声字幕、本人录音、授权配音还是字幕局部修复。原声优先复用;用户指定声音、服务、音乐时保持,不默认克隆声音或更换服务。

## 音频与正文
锁定最新正文。新的 TTS 读取使用者的音色授权、服务/模型、表达和音频参数,凭据由现有环境提供。表达指令与正文分离,防止被朗读;无法消费要求的工具明确待配音,不静默换声或接口。
分段保持声音状态,开场好奇、解释清楚、转折有力等仅按内容与用户风格使用,不固定夸张情绪。缓存绑定完整正文、模型、声音、表达与音频参数;结果不明先恢复原任务,避免重复付费。

## 时间与字幕
工具文件:[check_srt.py](scripts/check_srt.py),随技能目录一起安装。
测量实际音频后,按本次时间戳或本地转写建立时间线。ASR 是候选,回听名称、数字、外文、拼接、否定词与关键限定;权威写法与原意冲突时标注,不暗改说话者事实。
按语义分段,保留 UTF-8 软字幕。先将 `skill_dir` 定位为本 `SKILL.md` 所在目录,再从任意工作目录运行 `python "<skill_dir>/scripts/check_srt.py" "<字幕绝对路径.srt>" --duration <实际时长秒>`,可配置 `--max-cps`、`--min-duration`;格式/越界为错误,阅读速度与交叠为待观察提示。多说话者交叠可能合理,严格模式按需求使用。
使用 [音频字幕检查](references/audio-captions.md) 记录单词时间与句子时间的粒度、边界和未知区间。最终字幕不使用稿件估时。声音改变要同步重建受影响的分镜、动画事件、字幕和章节,不靠硬加速掩盖不足。

## 实际审阅
烧录是可选工序,另存版本,保留软字幕。字幕在当前画幅和手机大小检查换行、安全区、遮挡与原有画面文字;不要把固定字号当通用可读标准。
响度、峰值与波形只做技术检查。试听实际文件的发音、语气、停顿、段落接缝、音乐与人声关系;需要音乐时核验具体曲目使用权,平滑首尾并保留分轨。
没有听辨或读图能力就写清待验收范围,不能仅凭转写、成功生成或工程参数签听感通过。

交付当前音频、SRT/按需ASS、来源与授权记录、实际时长、配置指纹和观察记录。私人声音引用与服务日志留在工作项目,不提交到公开仓库。

Files in this skill

  • SKILL.en.md2.9 KB
  • SKILL.md2.9 KB
  • agents/openai.yaml370 B
  • references/audio-captions.en.md1.2 KB
  • references/audio-captions.md1.1 KB
  • scripts/check_srt.py3.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.