horizontal-content-talk-video

name	horizontal-content-talk-video
description	横版口播内容增强剪辑。人物始终是主视觉，左右安全区放标题、步骤、对比、结果卡，组件克制不抢话。 适合横版口播课、工具教程、商业分析、方法论讲解、案例复盘。 当用户说"横版口播增强"、"横版讲解视频"、"横版内容增强剪辑"、"人物为主横版"时使用。

Horizontal Content Talk Video

Goal

Turn a 16:9 horizontal talking-head video into a polished explainer where the speaker remains the primary visual anchor. Components support understanding but never dominate the frame.

This skill is an editing standard. It produces a scene plan, caption timing, material placement, and quality checklist. If a Remotion template is available, use it for rendering.

Required Inputs

• Source talking-head video (16:9)
• Original script, if available
• Material folder, if assets are provided
• Title and CTA
• Output target: 1920x1080, 30fps, H.264/AAC

Workflow

1. Inspect source with ffprobe. Extract frames near 1s, 8s, and end to judge face position and safe zones.
2. Transcode HEVC or unstable phone footage to 1920x1080 H.264/AAC.
3. Transcribe audio with Whisper. Correct names, product terms, numbers, and CTA words. Check the final 10 seconds separately.
4. Segment script into 5-7 chapters. Each chapter 6-14 seconds.
5. For each chapter, decide layout: full speaker, speaker + side cards, statement card, comparison, CTA.
6. Assign materials to fixed slots before animating.
7. Create props or edit timeline with caption times, scene times, and phrase-triggered items.
8. Render stills for key sections before full render.
9. Render full video, generate contact sheet, verify quality gates.

Layout Rules

• Speaker face is the primary visual. Do not cover with components.
• Face safe zone: roughly 35%-65% width, 15%-65% height.
• Left/right info areas: title top-left, cards on alternating sides, never both sides simultaneously.
• Subtitle area: bottom center, fixed position, large font, 1-2 lines preferred.
• Do not stack more than 2 large components at the same time.
• Cards enter by phrase trigger, not by timer.

Component Patterns

• Hook: Opening question with contrast card. One big title, one small card.
• Pain: Cost/time/frustration cards. Use red or warning accent. Left title + right cards.
• Outcome: Pivot statement. Large text + result card. Bright accent.
• Steps: Numbered method cards. Vertical stack on one side, current step highlighted.
• Proof: Big number + supporting cards. Keep minimal.
• Compare: Side-by-side before/after cards. Moderate width, face-safe.
• CTA: Clean action card + synchronized subtitle. No early ending.

Animation Rules

• Component lead: 0.18s
• Caption lead: 0.06s
• Card reveal: ~8 frames
• Chapter title fade: ~12 frames
• One semantic beat gets one main animation only
• Do not let components lag behind speech — horizontal pacing feels slower than vertical

Caption Sync

• Use Whisper timestamps, then manually correct
• Never repeat previous sentence
• Final 10 seconds: check separately, most prone to lag
• Numbers and CTA must be exact
• Traditional Chinese → Simplified before rendering

Quality Gates

Fail if any of:

• Not 1920x1080
• Audio missing or duration mismatch
• Face blocked by components
• Captions lag or lead noticeably
• Components overlap hiding important content
• Template debug words visible (Hook, Action, Scene, version labels)
• No visible structure change across the video

Validation Commands

ffprobe -v error -show_entries format=duration,size,bit_rate -show_streams -of json "output.mp4"
ffmpeg -y -i "output.mp4" -vf "fps=1/5,scale=480:-1,tile=5x3" -frames:v 1 -update 1 "contact.jpg"

Reference

Read references/content-sop.md for the full SOP with detailed layout rules, component specs, timing heuristics, and delivery checklist.