本篇只聚焦一个类别：视频处理。以下 Prompts 用于让 AI 直接产出可运行/可构建/可部署的在线字幕工具项目（包含完整代码、文件结构、运行命令、部署说明，以及测试用例或 QA checklist）。
在线视频/音频字幕提取与 SRT 导出工具
上传视频或音频，自动转写并生成可下载的 SRT/VTT，同时保留时间轴与置信度信息。
英文 Prompt:
You are a senior full-stack engineer. Build a production-ready web app named "Sub Extractor".
Goal
- Users upload a video (mp4/mov/webm) or audio (mp3/wav/m4a).
- The app extracts audio, runs speech-to-text, and exports sub s.
Hard requirements
- Output: SRT and WebVTT downloads.
- Provide: full source code, complete file tree, and step-by-step run commands.
- Provide: Dockerfile + docker-compose for local deployment.
- Provide: a minimal REST API (upload -> job -> result) with job status polling.
- Provide: at least 8 test cases (unit/integration) OR a QA checklist with expected results.
Tech constraints
- Use Node.js 20 + Type .
- Frontend: Next.js (App Router) + Tailwind.
- Backend: Next.js route handlers OR separate Express service; choose one and justify.
- Use FFmpeg (server-side) to extract audio; do not rely on system-specific paths.
- For STT: implement a pluggable adapter interface; include a local "mock" provider and document how to switch to Whisper.
UX
- Drag-and-drop upload.
- Show progress: upload %, processing steps, ETA.
- Show a tran preview with timestamps.
Deliverables
1) File tree
2) All code files
3) Commands: install, dev, test, build
4) Docker deployment instructions
5) Tests/QA checklist
6) Security notes (file size limits, MIME validation, temp file cleanup)
中文释义:
让 AI 生成一个可部署的字幕提取站点：支持上传媒体、服务端用 FFmpeg 抽取音频、可替换的转写适配器、任务队列与状态轮询，并输出 SRT/VTT；同时给出 Docker 部署与测试/QA 清单，确保真实可用。
在线 SRT 时间轴校对与批量替换编辑器
针对已有字幕文件进行可视化修正：时间轴拖拽微调、批量查找替换、合并/拆分字幕段落。
英文 Prompt:
Build a web- d SRT editor.
Features
- Import/export: SRT + WebVTT.
- Parse, validate, and repair common issues: overlapping cues, invalid timestamps, missing indices.
- Timeline editor: shift selected cues by +/- ms, stretch/compress segment, snap to grid.
- Batch operations: find/replace, regex replace (safe mode), remove filler words list.
- Undo/redo history.
Tech stack
- React + Type + Vite.
- No backend needed for core editing; everything runs in-browser.
- Provide an optional Node API to save/load projects.
Deliverables
- Full file tree + complete code.
- Run commands.
- A set of 10 example SRT fixtures (as strings in tests) that cover edge cases.
- Unit tests for parser, serializer, and repair functions.
UI requirements
- Two-pane: left cue list, right cue editor.
- Keyboard shortcuts: J/K navigation, Ctrl+F search, Ctrl+Z undo.
中文释义:
让 AI 产出一个前端为主的在线字幕编辑器：重点是 SRT/VTT 的解析、修复与批处理能力，配合时间轴微调与撤销重做；并用测试夹具覆盖重叠、断行、时间戳非法等常见坑。
在线字幕翻译与双语对齐工具（SRT -> 双语 SRT）
把原字幕翻译成目标语言，并生成双语字幕（同时间轴，单行/双行可切换）。
英文 Prompt:
Create a full-stack web tool: "Bilingual Sub Aligner".
Input
- Upload SRT.
- Choose source language and target language.
Processing
- Translate each cue text using an LLM translation adapter (pluggable).
- Preserve timestamps exactly.
- Support two output modes:
1) Single cue with two lines (original + translated)
2) Two consecutive cues (original then translated) with same time window
Quality controls
- Glossary terms: user-defined list of forced translations.
- Max line length: automatically reflow lines.
- Detect and warn if translated text expands too much.
Deliverables
- Next.js + API routes + a simple job queue in-memory.
- Full project code + file tree.
- Commands + Docker.
- Provide at least 12 QA checklist items (encoding, RTL languages, punctuation, emoji, line breaks, long sentences, SRT BOM, etc.).
Do NOT
- Do not generate images or suggest image generation prompts.
- Keep it strictly sub text processing.
中文释义:
让 AI 输出一个双语字幕对齐站点：输入 SRT，调用可替换的翻译适配器生成译文，严格保留时间轴；同时提供术语表、换行控制、膨胀检测等质量开关，并给出 Docker 与 QA 校验点。
在线字幕一致性检查与术语纠错工具
用于团队字幕规范：统一人名/品牌名译法、标点风格、数字/单位格式，并生成可下载的修订版。
英文 Prompt:
Build a web app that audits sub consistency.
Inputs
- SRT upload
- A "style guide" JSON (user editable):
- preferred terms map
- punctuation rules
- number formatting
- profanity filter (optional)
Outputs
- A report (HTML + JSON) listing issues by cue index/time.
- A "fixed" SRT download.
Implementation
- Type library for parsing + rules engine.
- UI to toggle rules and preview diffs per cue.
- Add a diff viewer (before/after) with highlight.
Deliverables
- Monorepo with packages: parser, rules-engine, web.
- Tests for at least 15 rule cases.
- Explain how to add a new rule plugin.
中文释义:
让 AI 生成一个字幕质量审计工具：核心是“规则引擎 + 报告 + 可回写修订”，适合批量统一术语/标点/数字格式，并以差异高亮方式让编辑快速确认修改。
在线字幕与媒体时长校验工具（异常检测与修复建议）
检查字幕是否超出视频时长、是否存在空洞/重叠、是否出现过长显示时间，并给出一键修复建议。
英文 Prompt:
Create a tool that validates sub s against media duration.
Inputs
- Upload video file AND SRT
Checks
- Sub end time must not exceed video duration.
- Detect gaps > N seconds (configurable).
- Detect overlaps and extremely short cues.
- Reading speed estimate (chars/sec) and flag too-fast cues.
Fix suggestions
- Shift all cues by offset.
- Clamp cues to duration.
- Auto-merge cues that are too short.
Tech
- Backend uses FFmpeg/ffprobe to read duration.
- Frontend shows charts: timeline heatmap, warnings list.
Deliverables
- Full code, file tree, run commands.
- Docker.
- At least 10 tests for the validation logic and parsing.
Note
- Strictly text/video processing; no image generation features.
中文释义:
让 AI 产出一个“字幕 vs 视频”对账工具：用 ffprobe 获取时长，再做重叠/空洞/阅读速度等校验，提供偏移、截断、合并等修复建议；适合发布前批量做质量门禁。
在线字幕片段剪辑清单生成器（按关键词自动打点）
输入字幕与关键词集合，自动输出“精彩片段清单”（起止时间、命中关键词、上下文），便于后续剪辑或标注。
英文 Prompt:
Build a web tool: "Sub Highlight Finder".
Inputs
- SRT upload
- Keywords list (comma-separated) + optional regex mode
Output
- A CSV + JSON timeline of highlight segments:
- start/end time
- matched keyword(s)
- cue indices
- context text
Features
- Merge nearby hits into a single segment if within X seconds.
- Provide a preview player that seeks to the segment times (use an HTML5 video element).
Deliverables
- Full code (React + Node).
- Example dataset and tests for keyword matching + segment merging.
- Provide a CLI as well (node) to process files in batch.
中文释义:
让 AI 生成一个“字幕打点”在线工具：以 SRT 为输入，按关键词/正则自动合并相邻命中，输出 CSV/JSON 的片段清单，并提供可跳转预览；同时附带一个可批处理的 CLI，方便团队流水线使用。
在线字幕格式转换与编码修复工具（SRT/VTT/ASS）
解决跨平台字幕兼容性：转换格式、修复编码与换行、统一时间戳精度，并输出校验报告。
英文 Prompt:
Create a sub format converter web app.
Supported
- Input: SRT, WebVTT, ASS (subset)
- Output: SRT and WebVTT (required), ASS (optional)
Repair
- Detect encoding issues (UTF-8 BOM, Windows-1252) and normalize to UTF-8.
- Normalize line breaks (CRLF/LF).
- Normalize timestamp precision.
Deliverables
- Full code + file tree.
- Explain parsing limitations for ASS.
- Provide 20 fixture samples and tests.
- Provide a "lint" report output with warnings and suggested fixes.
中文释义:
让 AI 产出一个字幕格式转换站点：重点在“编码与格式修复 + 可解释的 lint 报告”，将输入统一到 UTF-8，并把常见兼容性问题（BOM、换行、时间戳精度）显式提示出来。
在线字幕项目打包与交付物生成器（文件树/命令/部署/QA 一键导出）
把字幕项目当作交付件：生成标准化项目结构、README、Docker、CI 配置与 QA 清单，便于团队交接。
英文 Prompt:
Build a web app that turns sub work into a reproducible deliverable package.
Input
- Upload one or more sub files (SRT/VTT) + optional media data JSON.
Output
- A downloadable zip containing:
- /sub s originals and normalized versions
- /reports lint report
- /tools a small Node to validate timestamps
- README with commands and environment requirements
- Dockerfile + docker-compose
- QA checklist markdown
Requirements
- The site should generate the zip on the server.
- Must include a deterministic file tree.
- Provide tests ensuring zip structure and deterministic ordering.
Deliverables
- Full code, file tree, commands, Docker.
- Security constraints: zip bomb prevention, size limits, temp cleanup.
中文释义:
让 AI 生成一个“字幕交付包生成器”：上传字幕后自动产出规范化文件树、报告、校验脚本、README、Docker 与 QA 清单，并以 zip 形式一键下载；适合把一次字幕处理工作变成可复现、可交接的项目交付物。
