这篇文章只围绕一个类别：音频处理。下面每条 Prompt 都要求 AI 直接产出一个可运行/可构建/可部署的在线工具（完整项目代码 + 文件结构 + 运行命令 + 部署说明 + 测试用例/QA checklist）。主题聚焦：播客/访谈音频的 响度标准化（LUFS） 与 静音段检测剪切（静音剪切=删除或标记现有音频中的静音区间，不是生成图片）。
浏览器端：FFmpeg.wasm + WebAudio 的响度标准化与静音剪切工具
适合无需上传到服务器、强调隐私的本地处理场景。
英文 Prompt:
Build a production-ready web app that runs fully in the browser to:
- Detect silence segments in an audio file (threshold + min duration).
- Normalize loudness to a target LUFS (e.g., -16 LUFS for podcasts) using EBU R128.
- Export WAV and MP3.
Tech stack:
- Vite + React + Type
- ffmpeg.wasm (in-browser)
- WebAudio API for preview
Hard requirements:
1) Output a complete project with file tree and all source code.
2) Provide exact install/run commands.
3) Provide a clear UI spec: upload, waveform/scrub preview, parameters, job progress, download results.
4) Provide an offline-friendly mode (service worker optional) and explain limitations.
5) Provide at least 8 QA checks + 5 automated tests (Vitest) covering: silence detection accuracy, LUFS target, export integrity, edge cases (short clips, stereo/mono, VBR mp3).
6) Include a README with deployment steps (static hosting) and troubleshooting.
Do NOT generate images. Do NOT mention Midjourney/Stable Diffusion.
中文释义: 让 AI 直接交付一个“纯前端”的音频处理工具项目，包含完整代码、运行命令、部署与测试，能做静音段检测剪切与 LUFS 标准化并导出。
服务端队列：Next.js + BullMQ + FFmpeg 的批量处理工作台
适合需要同时处理多文件、并发与任务监控的团队场景。
英文 Prompt:
Create an online batch audio processing tool with:
- Multi-file upload (drag & drop), per-file status, retry, cancel.
- Pipeline: analyze loudness -> normalize to target LUFS -> silence detect -> trim/keep with configurable rules -> export.
- Job queue + worker.
Tech stack:
- Next.js (App Router) + Type
- Node.js worker using ffmpeg (native) + fluent-ffmpeg
- BullMQ + Redis for queue
- SQLite (or Postgres) for job data
Deliverables:
- Full repo code + file tree.
- Docker Compose for app + redis.
- CLI s for local dev.
- API routes (upload, create job, job status, download).
- Security: rate limit, file size limits, safe temp storage cleanup.
- Minimum 10-item QA checklist + at least 6 automated tests (Jest) including: queue behavior, parameter validation, download auth, cleanup.
No image generation prompts.
中文释义: 让 AI 输出一个带队列/Worker 的在线“音频批处理工作台”，重点是可运行、可部署、可测试，以及任务可视化与安全限制。
Python 后端：FastAPI + Celery 的响度分析与标准化 API + Web UI
适合想用 Python 生态快速扩展分析报告的场景。
英文 Prompt:
Design and implement a web tool that provides:
- Upload audio -> compute loudness metrics (integrated LUFS, true peak, loudness range).
- Normalize audio to target LUFS with limiter.
- Silence detection with a report: segments list (start/end), total silence duration.
- Export processed audio and JSON report.
Tech stack:
- FastAPI + Python 3.12
- Celery + Redis
- FFmpeg + ffprobe
- Simple frontend (HTMX or React) with job polling
Must include:
- Full source code and file tree.
- Commands to run locally and with Docker.
- Data schema for job tracking (SQLite ok).
- At least 5 test cases (pytest) + a QA checklist.
- Clear deployment guide (Docker) and security notes (upload validation, path traversal prevention, temp cleanup).
No image generation content.
中文释义: 让 AI 产出“音频响度分析 + 标准化 + 静音报告”的在线工具，强调 API + UI、可部署与测试。
参数可解释：面向播客的预设与一键处理（-16 LUFS / -19 LUFS）
适合非技术用户，降低参数门槛，提供明确预设。
英文 Prompt:
Build a web app focused on podcast presets:
- Presets: Podcast stereo -16 LUFS, Podcast mono -19 LUFS, Audiobook -18 LUFS.
- Advanced settings hidden behind a toggle (silence threshold, min silence, padding).
- Before/after A/B preview (30s snippet) and a final export.
Stack:
- Remix or Next.js + Type
- Backend FFmpeg processing
Requirements:
- Provide full code + file tree.
- Provide exact run commands.
- Include UX copy (Chinese UI ok) explaining what LUFS means and when to pick each preset.
- Add at least 8 QA checks and 5 automated tests.
- Add a "dry run" mode that outputs only the analysis report without modifying audio.
No image generation.
中文释义: 让 AI 交付一个带“播客预设”的在线工具，提供一键 LUFS 标准化 + 静音剪切，并支持干跑输出分析报告。
静音剪切策略：删除、保留、或只做标记（生成时间轴报告）
适合编辑工作流：先做“报告/标记”，再决定是否剪切。
英文 Prompt:
Create an online tool that supports 3 silence handling modes:
1) Remove silence segments.
2) Keep silence but lower volume (ducking).
3) Keep audio unchanged and generate an edit decision list (EDL/JSON) with silence segments.
Include:
- A timeline UI with segments list and timestamps.
- Export JSON + CSV reports.
- FFmpeg commands that implement each mode.
Deliver:
- Full project code, file tree, commands.
- Docker deployment.
- Minimum 10-item QA checklist + at least 6 automated tests.
No image generation instructions.
中文释义: 让 AI 输出一个“静音处理策略可选”的在线工具，既能剪切也能只生成时间轴报告，便于接入剪辑流程。
批量命名与打包：处理结果自动打包 ZIP 并附带报告
适合一次处理多集播客或多段采访录音的交付场景。
英文 Prompt:
Build a batch processing web tool:
- Upload multiple files.
- Configure naming template: {original}-{lufs}-{mode}.mp3
- After processing, generate a ZIP containing all outputs plus per-file JSON reports.
Stack:
- Node.js backend + FFmpeg
- Frontend with progress per file
Must include:
- Full code + file tree.
- Commands to run.
- Storage strategy (local disk temp) with automatic cleanup.
- Security checks (file type validation, size limits).
- At least 5 automated tests + 10 QA checklist items.
No image generation.
中文释义: 让 AI 交付一个面向“批量交付”的在线工具：批量处理、统一命名规则、最终打包 ZIP，并输出报告与测试。
导出格式：WAV/MP3/AAC 互转 + 元数据保留（标题/作者/封面不处理）
适合需要标准化输出格式与元数据的发布流程。
英文 Prompt:
Create a web app that processes audio and exports to WAV/MP3/AAC.
- Preserve basic audio data tags where possible ( /artist/album).
- Do NOT generate or modify cover images; ignore artwork fields if present.
- Provide a clear explanation of what data is preserved per format.
Include:
- Full code + file tree.
- Run/deploy commands.
- QA checklist (at least 10 items) and automated tests (at least 5).
- Edge cases: VBR MP3, different sample rates, mono/stereo conversions, clipping prevention.
中文释义: 让 AI 输出一个“格式转换 + 响度/静音处理”的在线工具，并明确元数据处理边界（不做封面图相关操作）。
可测试交付：附带 Playwright 端到端用例与可复现测试音频生成脚本
适合团队协作与持续集成，保证工具行为稳定可回归。
英文 Prompt:
Enhance the web tool by adding a robust test suite:
- Provide Playwright E2E tests that cover: upload, parameter changes, job creation, progress UI, download.
- Provide s to generate deterministic test audio (tones + silence) using FFmpeg so tests are reproducible.
Requirements:
- Full code additions with file tree updates.
- CI workflow (GitHub Actions) running unit + e2e tests.
- At least 8 QA checklist items.
No image generation prompts.
中文释义: 让 AI 给在线工具补齐自动化测试与可复现测试音频生成脚本，便于 CI 跑回归。
