playbooks

home / skills / cklxx / elephant.ai / video-production

video-production skill

/skills/video-production

This skill guides end-to-end video production from brief to delivery, ensuring reproducible workflows, platform-ready exports, and traceable decisions.

npx playbooks add skill cklxx/elephant.ai --skill video-production

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
3.1 KB
---
name: video-production
description: 端到端视频制作 playbook（策划→素材规范→剪辑→字幕→导出→发布→验收），强调可复现的交付流程。
---

# 视频制作（从 Brief 到交付）

## When to use this skill
- 需要把碎片素材（口播/B-roll/屏录/图片/字幕）剪成可验收的成片。  
- 需要适配多平台比例（16:9 / 9:16 / 1:1）并遵循导出规格。  
- 需要“可复现”的交付：素材清单、剪辑决策、导出参数、验收记录可追溯。

## 必备输入
1) **受众与目标**：希望观众做什么（了解/转化/培训/投放）？成功指标？  
2) **平台与比例**：YouTube / B 站 / TikTok / Reels / 内部培训；目标分辨率/帧率。  
3) **时长与结构**：开头 hook（前 3–10 秒）、主体、结尾 CTA；是否需要章节/时间轴。  
4) **品牌规范**：字体/色板/Logo/水印/片头片尾/字幕样式（字号/描边/阴影/高亮规则）。  
5) **音频要求**：是否降噪/配乐/旁白；响度与 true peak 目标（常用参考：-14 LUFS / -1 dBTP）。  
6) **交付物**：MP4 + 字幕（SRT/VTT）+ 封面图 + 变更说明；是否需要 fast start。

## 工作流
1. **预制作**：写脚本（旁白+强调词+停顿点）、镜头单（镜头类型/时长/所需素材）、资产清单（来源/授权）。  
2. **素材预检**：抽检分辨率、帧率、像素宽高比、时长、编码器/码率、音频采样率/声道、是否 VFR。  
3. **粗剪 → 精剪**：先跑通结构，再控节奏与信息密度（口播“可理解、可跟上、可复述”）。  
4. **声音优先**：人声明晰优先于画面高级感；配乐给人声留空间。  
5. **画面统一**：最小必要调色（统一曝光/白平衡/对比度）；避免每段素材像不同片子。  
6. **字幕与可访问性**：移动端优先（字号、对比度、安全边距）；关键术语一致；公开视频建议同时交付“烧录版+独立字幕文件版”。  
7. **导出**（YouTube SDR 参考）：容器 MP4；视频 H.264 progressive/high profile（常见 4:2:0）；音频 AAC 48kHz；fast start；1080p 约 8 Mbps（24/25/30fps）或 12 Mbps（48/50/60fps）；4K 约 35–68 Mbps；响度 -14 LUFS，true peak ≤ -1 dBTP。  
8. **发布与验收**：逐条勾验（画面无黑帧/比例正确、声音无爆音且响度达标、字幕对齐、导出帧率/码率符合平台、fast start 生效），记录变更与已知限制。

## 本仓库工具链说明
- 当前仓库不提供内建视频处理工具链；请根据团队规范选择并记录使用的外部工具与版本。

## 参考资料
- YouTube 上传建议：https://support.google.com/youtube/answer/1722171  
- Adobe 导出参数权衡：https://www.adobe.com/creativecloud/video/hub/guides/best-export-settings-for-premiere-pro.html  
- APU：YouTube 响度目标（-14 LUFS / -1 dBTP）：https://apu.software/youtube-audio-loudness-target/  
- APU：常见平台响度目标表（含 EBU R128）：https://apu.software/loudness-standards/  
- W3C WAI：字幕（Prerecorded）资源：https://www.w3.org/WAI/WCAG22/Understanding/captions-prerecorded.html

Overview

This skill is an end-to-end video production playbook that turns briefs and fragmented assets into reproducible, verifiable deliverables. It codifies inputs, a stepwise workflow (planning → assets → edit → captions → export → publish → acceptance), and concrete export and loudness targets so outputs are consistent across platforms. The emphasis is traceable delivery: material lists, edit decisions, export settings, and acceptance records are all reproducible and auditable.

How this skill works

The playbook guides teams through preproduction (script, shot list, asset inventory), automated and manual asset inspection (resolution, frame rate, codec, audio sample rate, VFR checks), and staged editing (rough cut → fine cut) with audio-first priorities. It specifies minimal grading, subtitle standards, export parameters for common platforms, and a checklist-driven publish and acceptance process that records deviations and known limitations.

When to use it

  • Assemble short- or long-form videos from mixed source material (talking head, B-roll, screen capture, images).
  • Deliver platform-specific masters and variants (16:9, 9:16, 1:1) with verified export specs.
  • Require reproducible deliveries where asset manifests, edit rationales, export params, and acceptance logs must be traceable.
  • Prepare accessible releases with burned-in captions and separate subtitle files (SRT/VTT).
  • Enforce consistent audio loudness and technical compliance for ad or platform submission.

Best practices

  • Start with clear audience, goal, platform, and success metrics; define target resolution and frame rate up front.
  • Perform a technical precheck on every clip: resolution, pixel aspect, codec, VFR, audio sample rate, channels.
  • Prioritize voice clarity over visual polish; leave space for dialogue in the mix and set loudness to -14 LUFS, true peak ≤ -1 dBTP.
  • Keep color work minimal—unify exposure and white balance to avoid pieces looking like different projects.
  • Deliver both burned-in captions for social previews and separate SRT/VTT files for accessibility and platform uploads.
  • Use a checklist for export and acceptance: no black frames, correct aspect, codec/bitrate targets, fast start enabled, caption sync.

Example use cases

  • Convert a collection of recorded lectures and slides into a branded training series with chapter markers and subtitles.
  • Produce short marketing clips from interview footage and B-roll, outputting 16:9 masters plus 9:16 social edits.
  • Aggregate remote creator clips with varying codecs and frame rates, perform standardization checks, and produce a unified episode.
  • Prepare ad creatives with platform-specific bitrate and loudness settings and record acceptance evidence for media buys.
  • Deliver internal how-to videos with burned-in safety captions and separate VTT files for LMS ingestion.

FAQ

Which export settings should I use for YouTube?

Use MP4 container, H.264 progressive/high profile (4:2:0), AAC audio at 48 kHz, fast start enabled; 1080p ~8 Mbps for 24–30fps, 12 Mbps for 48–60fps, 4K ~35–68 Mbps; loudness -14 LUFS, true peak ≤ -1 dBTP.

What inputs are mandatory to start production?

Define audience and goal, target platform and aspect ratio, duration and structure (hook/body/CTA), brand specs (fonts/colors/logo/subtitle style), audio requirements, and expected deliverables.

Do you provide a built-in video processing toolchain?

No. The playbook does not include an internal processing toolchain—record which external tools and exact versions the team uses and document them in the delivery record.