视频处理 视频抽帧与时间轴标注 AI 提示词 (Prompts)

英文 Prompt:

You are to generate a complete, runnable web app that extracts existing frames from a user-uploaded video (NOT generating new images). Build a single-page app using Vite + React + Type , and use ffmpeg.wasm in the browser to seek and extract frames at specified timestamps. Features: upload video, show timeline with scrubbing, add labeled markers (timestamp + label + notes), extract a single frame at a marker, batch extract frames for all markers, show extracted frame thumbnails, export annotations as JSON, export frames + JSON as a ZIP. Output requirements: provide full source code with file tree, README with setup commands, build commands, and deployment to static hosting; include performance notes and fallback when wasm is slow; include at least 8 QA checklist items and 6 test cases (unit/integration) that validate marker CRUD, timestamp parsing, export correctness, and that no AI image generation is used.

中文释义: 生成一个可离线部署的前端工具：在浏览器用 ffmpeg.wasm 从视频里按时间点提取“已有帧”，并在时间轴上做标注、批量导出帧与 JSON 注释。

英文 Prompt:

Generate a full-stack online tool for extracting existing frames from videos and managing timeline annotations. Tech stack: Node.js (Fastify), Type , BullMQ (Redis) for job queue, and a React frontend. Users upload a video; server stores it, creates a job to extract frames at given timestamps using native ffmpeg, and returns progress via SSE. Features: create project, upload video, add/edit/delete markers, run extraction, view thumbnails, download a ZIP containing frames and annotations.json, and download a CSV of markers. Must include: file tree, complete source, env config, Docker Compose (app + redis), commands to run dev/prod, and deployment guide. Add security basics (upload size limits, path traversal prevention). Provide at least 10 QA checklist items and 8 API tests (e.g., using Vitest + supertest) for upload, marker endpoints, and export integrity. Clarify explicitly: do not generate images; only extract frames from the uploaded video.

中文释义: 生成一个带队列与进度的全栈工具：上传视频后由后端 ffmpeg 异步抽帧，前端可标注时间点并导出 ZIP/CSV/JSON，适合大文件与批量任务。

英文 Prompt:

Create a web- d timeline annotation tool focused on speed: users can load a video (local file) and add markers while playing using keyboard shortcuts. Stack: Next.js + Type . Requirements: playback controls, adjustable skip (e.g., 1s/5s/10s), hotkeys to add marker, edit label, categorize markers, and auto-capture current timestamp. Export: chapters JSON, CSV, and a YouTube-style timestamps text. Optional: if browser APIs allow, generate a small preview thumbnail by capturing from the video element (canvas drawImage) but only from the existing frame (no generative AI). Provide complete code + file tree + run/build commands + deployment. Include at least 6 test cases for parsing/formatting exports, hotkeys, and marker ordering, plus a QA checklist for cross-browser behavior.

中文释义: 生成一个更偏“标注效率”的在线工具：用快捷键在播放中打时间点，导出章节/字幕风格时间戳文本；可选从 video 元素抓取当前帧缩略图（仍是提取现有帧）。

英文 Prompt:

Build an online tool that helps users generate timestamp lists for frame extraction. Category: video processing utilities. Stack: Vue 3 + Type . Input: video duration (manual or parsed from uploaded file data), start/end time, step size, and named segments. The app generates a validated timestamp list (hh:mm:ss.ms), allows grouping, de-duplication, sorting, and merging. Output: a JSON config compatible with ffmpeg extraction, a plain text list, and a ready-to-run command snippet for ffmpeg (but do not execute it). Add a marker preview table and validation errors. Deliverables: full code, file tree, README with commands, deploy guide, and at least 7 unit tests validating timestamp math and formatting. Include QA checklist items for edge cases (overlaps, negative times, end boundary).

中文释义: 生成一个“批量时间点生成器”在线工具：按区间/步长/分段规则产出抽帧时间点列表与配置，并导出可直接用的 ffmpeg 命令片段（只生成文本，不执行）。

英文 Prompt:

Generate a web app that validates extracted frame sets against a marker list. Stack: SvelteKit + Type . Users upload (1) an annotations JSON/CSV and (2) a ZIP of extracted frames. The app checks: missing frames, extra frames, naming convention, timestamp rounding tolerance, and produces a report (HTML + JSON) with actionable fixes. Provide UI to configure naming templates (e.g., frame_{index}_{timestamp}.jpg). Include full source, file tree, dev/build commands, and deployment steps. Add at least 8 test cases for parsing CSV/JSON, zip listing, and report generation, plus QA checklist. Explicitly state: no image generation; only validation and reporting.

中文释义: 生成一个“抽帧结果校验/对账”在线工具：上传标注文件与抽帧 ZIP 后自动检查缺失/多余/命名不一致，并输出可读报告与修复建议。

英文 Prompt:

Build a backend service that creates a thumbnail sprite sheet and a WebVTT file for preview scrubbing. Stack: Python FastAPI + ffmpeg, with a minimal React UI to upload a video and download outputs. Requirements: configurable interval (e.g., every N seconds), max width, JPEG quality; output sprite.jpg and preview.vtt mapping times to sprite coordinates; include an HTML demo page that uses the VTT to show preview thumbnails on hover (scrubber). Provide full code, file tree, dockerfile, commands, deployment, and at least 6 automated tests for API endpoints and parameter validation. Make it clear: thumbnails come from existing video frames only; do not generate new imagery.

中文释义: 生成一个后端服务：从视频现有帧按间隔抽取缩略图，拼成雪碧图并生成 VTT 索引，前端示例可在拖动进度条时预览。

英文 Prompt:

Create an online timeline annotation tool that exports marker data into editing-friendly formats. Stack: Angular + Type . Users load a video, create markers and segments (in/out ranges), assign categories, and then export: (1) JSON, (2) CSV, (3) EDL text, and (4) Final Cut Pro (FC ) with markers or ranges. Provide full project code, file tree, commands, and deployment instructions. Include at least 7 unit tests for timecode conversion, EDL formatting, and generation. Add a QA checklist that verifies imports into common editors and handles FPS selection (e.g., 24/25/30).

中文释义: 生成一个标注工具：把时间点/区间导出成 EDL/FC 等剪辑工程友好格式，降低“标注到剪辑”的交付摩擦。

英文 Prompt:

Build a browser- d batch job plan generator for video operations. Category: video processing / batch processing. Stack: Vanilla Type + Tailwind (no framework). Inputs: a marker list (JSON/CSV), job presets (extract frames, extract audio, remux, create clips by segments), and output target paths. The app outputs: a bash and a PowerShell with ffmpeg commands for each job, plus a manifest JSON. Requirements: quoting/escaping, cross-platform path handling, dry-run mode (comments), and warnings for unsafe paths. Provide full source, file tree, README, and at least 8 tests for generation and escaping. Must include a QA checklist and explicitly avoid any generative image content; it only generates command s and manifests.

中文释义: 生成一个“批处理脚本生成器”在线工具：基于标注与预设输出 ffmpeg 脚本与清单，方便把抽帧/切片/封装任务交给 CI 或同事执行（工具只输出脚本文本）。