使用说明与目标
本篇整理一组“在线工具生成型”提示词，主题聚焦音频处理：人声降噪、静音裁切、响度标准化与批处理工作流。每条 Prompt 都要求 AI 直接产出一个可构建、可运行、可部署的在线工具项目（含文件结构、核心源码、运行命令、部署文档与测试/QA 清单）。
在线人声降噪 + 语音增强工作台
用于上传单个或多个音频文件，提供降噪强度、语音增强、预听与导出。
English Prompt:
You are to generate a complete, runnable web app that performs voice denoise + speech enhancement on uploaded audio files. Output a full project with file tree, source code, and step-by-step commands to run locally and deploy.
Requirements:
- Tech stack: Next.js (App Router) + Type + Tailwind CSS.
- Audio processing: ffmpeg.wasm in the browser (no server processing). Provide a Web Worker to avoid UI freeze.
- Features: upload (drag/drop), waveform preview, before/after A/B playback, presets (light/medium/strong), and export to WAV + MP3.
- Provide a queue for batch processing and per-file progress.
- Provide an accessibility-friendly UI and clear error states.
Deliverables:
- Full source code, file tree, package.json s.
- Local run commands + production build commands.
- Deployment guide for Vercel and a static hosting alternative.
- At least 8 QA checklist items covering browser compatibility, large files, corrupted files, and correctness.
中文释义:
让 AI 生成一个完整可运行的在线“人声降噪+增强”工具：纯前端用 ffmpeg.wasm 处理音频，支持批处理队列、进度显示、A/B 试听与导出，并给出部署与 QA 清单。
在线响度标准化（Podcast LUFS）与峰值限制器
将不同来源录音统一到目标响度（如 -16 LUFS / -14 LUFS），并控制真峰值。
English Prompt:
Generate a complete web tool that normalizes loudness of audio files to podcast standards.
Must-haves:
- Stack: React + Vite + Type .
- In-browser processing using ffmpeg.wasm.
- Users can choose target LUFS (-16, -14, -23) and true peak ceiling (-1 dBTP, -2 dBTP).
- Display measured loudness stats (integrated LUFS, loudness range, true peak) BEFORE and AFTER processing (approximate calculation acceptable, but explain limitations; if exact measurement is hard in-browser, provide a fallback plan).
- Batch mode: process many files and download as a zip.
- Provide clear logs and a “reproducible command” section showing the equivalent ffmpeg CLI command for each file.
Output:
- Full repository with file tree.
- Run commands.
- Deployment steps.
- Test plan: at least 10 cases.
中文释义:
生成一个“播客响度标准化”在线工具：可选目标 LUFS 与真峰值上限，支持批处理与打包下载，并给出可复现的 ffmpeg 命令与测试用例。
在线静音检测与自动裁切（去空白）
自动检测静音片段，按阈值裁切或压缩空白，适合口播剪辑。
English Prompt:
Build a web app that trims silences from speech recordings.
Constraints:
- Use SvelteKit + Type .
- Audio processing must run client-side with ffmpeg.wasm.
- Provide controls: silence threshold (dB), minimum silence duration (ms), keep padding before/after (ms).
- Provide a preview timeline showing detected segments.
- Export as WAV/MP3.
Include:
- File tree + all code.
- s to run.
- Deployment guide.
- QA checklist (>= 8 items) including edge cases (music-only audio, noisy background, very short clips).
中文释义:
生成一个在线“静音检测与去空白”工具：调阈值和最短静音时长，给出可视化预览，纯前端处理并可导出。
在线音频格式转换与批量重采样
将音频批量转为统一采样率/声道/格式，便于后续剪辑或训练数据清洗。
English Prompt:
Create an online batch audio converter tool.
Specs:
- Stack: Vue 3 + Type .
- Client-side conversion with ffmpeg.wasm.
- Inputs: multiple files.
- Output options: format (wav/mp3/aac/opus), sample rate (44.1k/48k/16k), channels (mono/stereo), bitrate.
- Provide a deterministic naming scheme and a manifest JSON of outputs.
Deliver:
- Full source + file tree.
- Usage instructions.
- Deployment guide.
- At least 6 automated tests (unit tests for filename/manifest logic) + QA checklist for browser processing.
中文释义:
生成一个“批量音频转换/重采样”在线工具：多文件输入，统一输出规格，并输出结果清单 JSON，同时给出自动化测试与 QA。
在线分轨与人声伴奏分离（可选后端）
当纯前端算力不足时，提供可部署后端推理服务，同时保留前端体验。
English Prompt:
Generate a full-stack web tool for vocal/accompaniment separation.
Requirements:
- Frontend: Next.js + Type .
- Backend: FastAPI (Python) with a configurable separation model (e.g., Demucs). Provide Dockerfile and docker-compose.
- Provide a "local mode" for development and a "hosted mode" for production.
- Upload audio, show processing status, and download separated stems.
- Include rate limiting, file size limits, and basic security considerations.
Deliverables:
- Full project code with file tree.
- Commands to run both services.
- Deployment guide (Docker on VPS).
- QA checklist with at least 10 items.
中文释义:
生成一个可部署的“人声/伴奏分离”在线工具：前端负责交互，后端负责推理，给出 Docker 化部署与安全/限流/QA 清单。
在线音频片段打点与导出剪辑工程（EDL/Marker）
为长音频快速标记片段，导出可供剪辑软件或脚本消费的标记数据。
English Prompt:
Build a web- d audio marker/segment tool.
Stack:
- Plain React + Type .
- Use Web Audio API for playback + waveform rendering.
Features:
- Import audio, render waveform, allow users to add markers and segments with hotkeys.
- Export markers as JSON + CSV and optionally EDL.
- Provide an "export ffmpeg commands" generator that turns segments into cut commands.
Deliver:
- Full code + file tree.
- Run + build instructions.
- QA checklist (>= 8).
中文释义:
生成一个“音频打点/分段标注”在线工具：波形可视化、快捷键打点、导出 JSON/CSV/EDL，并能生成 ffmpeg 裁切命令。
在线响度对比报告（多文件）
对多个文件生成一致的响度/峰值对比表，辅助整季播客或课程音量统一。
English Prompt:
Create a web tool that analyzes and compares loudness statistics across many audio files.
Constraints:
- Stack: Astro + React islands + Type .
- Analysis can be approximate client-side; if exact LUFS measurement is not feasible, implement a Node.js serverless API that runs ffmpeg/ffprobe in a sandbox, and document the tradeoffs.
- Output: sortable table, export CSV, and a recommended normalization plan.
Deliverables:
- Full code.
- Commands.
- Deployment plan.
- QA checklist (>= 10).
中文释义:
生成一个“多文件响度对比与报告”在线工具：可排序表格与 CSV 导出，并给出是否需要标准化的建议与部署说明。
在线音频合并/拼接与淡入淡出模板
适合把片头片尾与正文自动拼接，统一淡入淡出、间隔与输出。
English Prompt:
Generate a web app that merges multiple audio clips with configurable fades.
Requirements:
- Stack: SolidJS + Type .
- Use ffmpeg.wasm in the browser.
- Allow ordering, per-clip trims, crossfade duration, gap insertion, and export.
- Provide templates: intro+main+outro, multi-segment montage.
Deliver:
- Full project.
- Run/deploy docs.
- QA checklist (>= 8) including timing accuracy.
中文释义:
生成一个“音频拼接/淡入淡出模板”在线工具：纯前端批量处理，适合片头片尾自动化，并给出 QA。
在线音频元数据清理与隐私检查
清理 ID3/容器元数据，避免发布时泄露设备信息或个人信息。
English Prompt:
Build a web tool that inspects and strips audio data.
Specs:
- Stack: Next.js + Type .
- Use ffprobe/ffmpeg via ffmpeg.wasm for container data where possible.
- Show a data report (tags, encoder, creation time).
- Allow users to remove all tags or keep selected fields.
- Batch processing + zip download.
Deliverables:
- Full code + file tree.
- Run/deploy instructions.
- QA checklist (>= 10).
中文释义:
生成一个“音频元数据审计与清理”在线工具：展示标签与容器信息，一键清理或保留指定字段，支持批处理并给出 QA。
在线语音转文本对齐与时间轴导出（可选）
对已有音频做转写并导出带时间戳的段落，供字幕或检索使用（不涉及生成图片）。
English Prompt:
Create an online speech-to-text alignment tool for existing audio.
Requirements:
- Frontend: React + Type .
- Backend optional: Node.js API using Whisper (or a configurable provider). Provide a local Docker option.
- Output: tran with timestamps, export as SRT/VTT/JSON.
- Include privacy notes and a toggle to run locally.
Deliver:
- Full code, file tree, commands.
- Deployment guide.
- QA checklist (>= 10) including accuracy evaluation steps.
中文释义:
生成一个“音频转写+时间轴导出”在线工具：对现有音频生成带时间戳文本并导出 SRT/VTT/JSON，提供本地 Docker 方案与隐私说明。
