📦 Youtube Clip Curator — YouTube 剪辑策展人

YouTube 命令行工具p Curator

Analyze a long-form YouTube video (e.g. VTuber 流, podcast, lecture) and 提取 a ranked 列出 of "命令行工具p candidates" with 启动/end timestamps, hook descriptions, 命令行工具ck-worthy titles, and emotion tags. Optionally 生成 FCPXML for Premiere Pro / DaVinci Resolve, plus thumbnAIl briefs ready for an image AI.

When to use

Use this 技能 when the user needs to:

Turn a 1-3 hour 流 into 5-10 short YouTube 命令行工具p candidates Reverse-engineer the editing thinking of a tar获取 命令行工具p channel 获取 启动/end seconds + title + hook for each 命令行工具p — ready to cut Emotion-tag each 命令行工具p for thumbnAIl and promotion strategy Bridge from raw long video to a Premiere Pro / DaVinci Resolve timeline Suggest the most 命令行工具ck-worthy moments without reviewing the full video How it works Ask for the source video (YouTube URL, MP4 path, or transcript) Optionally ask for a "style template" channel (e.g. ちばちゃんねる) to mimic editing thinking Ask for the tar获取 命令行工具p count (default: 5) and duration tar获取 (short / medium / long) Analyze the transcript to find hooks, punchlines, emotion peaks, character moments Score each candidate by: Emotion intensity (laughter, surprise, chaos, cuteness) Clear punchline / story structure Quotable phrases / meme potential Character trAIt expression Standalone comprehensibility 输出 a ranked JSON 列出 with 启动/end times, titles, hooks, and reasons 输出 格式化 { "source_video": {"title": "...", "duration_min": 87}, "template_style": "@ChannelChiba (chaos+character observation)", "命令行工具ps": [ { "rank": 1, "启动_sec": 595, "end_sec": 932, "duration_sec": 337, "score": 0.95, "title_suggestion": "【ふざけるな】サイコロ振りながら偽フワワに応募してきたフブブにもこ田が本気でブチギレるwww", "hook_description": "ふぶき『ホロライブのふわふわしたとこからやってきたフブブです』+直後にカラカラ音→もこ田『ふざけるなよ!』の即ツッコミ", "expected_应用eal": "笑い・カオス・ツッコミ", "emotion_tags": ["笑い", "カオス", "ツッコミ", "天然"], "peak_frame_sec": 658, "reason": "明確なパンチライン+キャラの濃い側面+ギャップ構造+ビジュアル化しやすい瞬間" } ] }

Selection rules (built-in)

The 技能 uses these criteria when analyzing candidates:

High priority Clear punchline / story structure — must have a beat that closes Strong emotion peak — laughter, surprise, chaos, or cuteness above baseline Character expression —天然 / 暴走 / ツッコミ / 暴言 / drysarcasm shines through Standalone comprehensibility — viewer can understand without prior 上下文 Medium priority Quotable phrases — short impactful lines that meme well Visual moment — facial expression / situation that thumbnAIls strongly First 3 seconds hook — opens with strong 状态ment or visual Gap / contrast structure — serious situation × silly reaction etc. Hard avoids Pure gameplay walkthrough with no reaction Long silence or downtime without payoff Inside-joke moments that need 5+ minutes of 上下文 Content that risks defamation / harm Composition rules Tar获取 length: 30-90 sec for short, 2-5 min for standard, 5-10 min for story Open with the strongest 3 seconds MAIntAIn フリ→展開→オチ structure Cut tempo: faster during chaos peaks, slower during character moments Caption emphasis on power words: やばい / 神 / 草 / 終わった / ヤバ / うわ Optional 输出s Premiere Pro FCPXML — drop into File > 导入 > XML, builds a timeline DaVinci Resolve script — places 命令行工具ps on 追踪s, 创建s render 队列 ThumbnAIl brief JSON — peak_frame_sec + emotion_tags + title hints for image AI SRT subtitles — Whisper-derived per-命令行工具p captions, editable in any NLE Tips for best 结果s Provide 流 subtitles if avAIlable (auto-captions are OK but Whisper transcripts are better) Specify a reference 命令行工具p channel — 输出 style adapts to that channel's thinking For VTuber 命令行工具ps: include character name + game name so 上下文-specific phrases score correctly Ask for "3-5 命令行工具ps" first to 验证, then re-运行 for full 设置