youtube-seo-video

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Single Video Deep Analysis (Advanced)

单视频深度分析(高级版)

Diagnose why a video underperforms and produce a paste-ready fix kit. Retention and CTR are the primary levers; metadata is secondary. Always ask for the Studio CSV export if the user wants retention diagnosis — never guess Tier 1 numbers.
诊断视频表现不佳的根本原因,并生成即贴即用的修复包。 留存率和CTR是核心杠杆,元数据次之。如果用户需要留存诊断, 务必要求提供YouTube Studio导出的CSV文件——绝对不要臆测一级指标数值。

Data Collection (in order)

数据收集(按顺序)

  1. WebFetch the watch URL → title, description (partial), view/like, duration, upload date, channel, visible chapters
  2. yt-dlp (
    scripts/fetch_video.py
    or direct call) → full JSON metadata, tags, full description, chapters, thumbnails, captions
  3. YouTube Data API (if key) →
    videos.list?part=snippet,statistics, contentDetails,topicDetails,status,player,liveStreamingDetails
    and
    captions.list
    +
    captions.download
    (if OAuth)
  4. Transcript: prefer manual captions > auto-captions > Whisper transcription from audio stream
  5. Studio CSV (user-provided): retention curve, traffic sources, CTR by source, impressions, audience tab, real-time first-24h curve
  6. Audio loudness:
    ffmpeg -i audio.m4a -af loudnorm=print_format=json -f null -
    → integrated LUFS, true peak
  7. Thumbnail file: download for analysis via
    scripts/analyze_thumbnail.py
  8. SERP grid: fetch top-10 for primary keyword, save competitor thumbnails and titles for differentiation scoring
  1. WebFetch 抓取观看页面URL → 获取标题、(部分)描述、观看/点赞数、 时长、上传日期、频道、可见章节
  2. yt-dlp(通过
    scripts/fetch_video.py
    或直接调用)→ 获取完整JSON 元数据、标签、完整描述、章节、缩略图、字幕
  3. YouTube Data API(如有密钥)→ 调用
    videos.list?part=snippet,statistics, contentDetails,topicDetails,status,player,liveStreamingDetails
    , 以及
    captions.list
    +
    captions.download
    (如有OAuth权限)
  4. 文字稿:优先级为 人工字幕 > 自动字幕 > 通过Whisper从音频流转录
  5. Studio CSV(用户提供):留存曲线、流量来源、各来源CTR、 曝光量、受众标签页数据、前24小时实时曲线
  6. 音频响度:执行
    ffmpeg -i audio.m4a -af loudnorm=print_format=json -f null -
    → 获取整体LUFS、真实峰值
  7. 缩略图文件:下载后通过
    scripts/analyze_thumbnail.py
    进行分析
  8. SERP网格:抓取主关键词的前10名结果,保存竞品缩略图和标题用于差异化评分

Analysis Dimensions

分析维度

1. Retention Curve (highest weight when Studio data provided)

1. 留存曲线(提供Studio数据时权重最高)

Diagnose from the Studio retention curve:
  • 0-15s intro: target ≥70% still watching. Below → hook problem. Diagnose: weak first sentence, no payoff preview, long logo animation, re-introducing yourself ("Hey guys welcome back..."), asking to subscribe before value delivery.
  • 15-60s premise: target ≥60%. Below → premise unclear or mismatch with title/thumbnail.
  • Retention cliffs: any drop >10% in <5s is a structural issue. Map cliffs to transcript timestamps and identify:
    • Tangent / digression
    • Ad or sponsor break placed badly
    • Pacing death (long, slow exposition)
    • Promise broken (title said X, video now does Y)
    • Visual monotony (static shot >20s without B-roll)
  • Mid-video sustain: target curve slope ≥-0.3%/sec. Steeper = boring middle.
  • Ending: the last 30s often rises (re-watchers, end-screen hover). If it drops hard, end-screen is poorly placed or content feels done before the promise delivered.
  • Spike detection: peaks = rewatched moments = high-value moments. Reuse as chapter titles, thumbnails, Shorts clips.
  • APV vs niche median: score video as ×median. Flag anything <0.8×.
If the curve is not provided, state explicitly that retention diagnosis is unavailable and score only metadata, thumbnail, and engagement proxies.
基于Studio留存曲线进行诊断:
  • 0-15秒开篇:目标留存率≥70%。低于该值 → 钩子存在问题。 诊断方向:开篇第一句无力、没有收益预告、logo动画过长、 重复自我介绍(“嘿大家好欢迎回来…”)、在提供价值前就要求订阅。
  • 15-60秒主题铺垫:目标留存率≥60%。低于该值 → 主题不清晰, 或与标题/缩略图不匹配。
  • 留存断崖:任何5秒内跌幅超过10%的情况都属于结构性问题。将 断崖位置对应到文字稿时间戳,排查以下原因:
    • 跑题/偏离主线
    • 广告或赞助植入位置不当
    • 节奏拖沓(冗长缓慢的陈述)
    • 承诺未兑现(标题说的是X,视频内容却是Y)
    • 视觉单调(静态镜头超过20秒且无B-roll辅助画面)
  • 视频中段留存维持:目标曲线斜率≥-0.3%/秒。斜率越陡说明中段越枯燥。
  • 结尾表现:最后30秒留存通常会回升(回看用户、悬停等待结束画面的用户)。 如果大幅下跌,说明结束画面位置不佳,或内容在兑现承诺前就已结束。
  • 峰值检测:曲线峰值 = 被反复观看的片段 = 高价值时刻。 可复用为章节标题、缩略图、Shorts短视频素材。
  • APV与细分领域中位数对比:以中位数的倍数为视频打分。低于0.8倍的需标记。
如果未提供留存曲线,需明确说明无法进行留存诊断, 仅对元数据、缩略图和互动代理指标进行评分。

2. Title (target: 60-70 chars, max 100)

2. 标题(目标长度60-70字符,最多100字符)

  • Length (flag >70 truncation risk, critical >100)
  • Keyword placement: primary keyword in first 40 chars for Search; emotional driver in first 40 chars for Browse (which surface matters?)
  • Hook type: curiosity, number, contrarian, benefit, authority, fear. A video should commit to one dominant hook type that matches surface.
  • Emotional intensity: rate 1-5. Below 3 = passive titles that lose Browse impressions.
  • Clickability vs. deliverability: can the video content back up the title? If no, score down (long-term CTR decay).
  • Front-load uniqueness: avoid "My thoughts on", "A quick video about"
  • Case consistency with channel brand
  • 长度(超过70字符标记有截断风险,超过100字符为严重问题)
  • 关键词位置:搜索流量场景下主关键词需放在前40字符; 推荐浏览(Browse)场景下情感驱动点需放在前40字符(需判断哪种曝光场景更重要)
  • 钩子类型:好奇心型、数字型、反常识型、利益型、权威型、恐惧型。 视频应明确采用一种与曝光场景匹配的主导钩子类型。
  • 情感强度:按1-5分评级。低于3分属于被动型标题,会损失Browse推荐曝光。
  • 点击吸引力与内容兑现度:视频内容能否支撑标题承诺?如果不能则扣分 (会导致长期CTR下滑)。
  • 前置独特性:避免使用“我对…的看法”、“一个关于…的简短视频”这类开头
  • 大小写风格与频道品牌保持一致

3. Description

3. 视频描述

  • Above-the-fold (first 150 chars): MUST include primary keyword, restate title intent, give a click-for-more reason
  • Full length: 1,500-4,000 chars ideal. Flag <500 thin, >6,000 bloated
  • Entity coverage: list 8-15 Knowledge Graph entities the video should mention for topic classification. Check transcript + description for coverage. Missing entities = missed semantic relevance.
  • Secondary keywords: 3-5 naturally placed in lines 2-10
  • Chapters block: present, first =
    0:00
    , ≥3, each ≥10s, descriptive
  • Links: grouped and labeled (affiliate, social, related, resource)
  • CTA: subscribe + next-video suggestion + lead magnet
  • Hashtags: 3 meaningful at the bottom, first is strongest
  • FTC disclosure: flag missing if affiliate/sponsored
  • Timestamps for Key Moments rich result: correctly formatted (
    M:SS - label
    or
    MM:SS label
    ), eligible for Google Search carousel
  • 首屏内容(前150字符):必须包含主关键词,重述标题核心意图, 给出点击查看更多的理由
  • 总长度:理想长度为1500-4000字符。低于500字符标记为内容过薄, 高于6000字符标记为内容冗余
  • 实体覆盖度:列出8-15个视频为完成主题分类应当提及的 知识图谱(Knowledge Graph)实体。检查文字稿和描述中的覆盖情况。 缺失实体 = 错失语义相关性机会。
  • 次要关键词:3-5个,自然分布在第2-10行
  • 章节区块:需存在,首个章节为
    0:00
    ,章节数≥3, 每个章节时长≥10秒,标题具有描述性
  • 链接:分组并标注类型(联盟链接、社交账号、相关内容、资源链接)
  • CTA(行动号召):包含订阅 + 下一个视频推荐 + 引流诱饵
  • 话题标签:底部放置3个有意义的标签,第一个为最强相关标签
  • FTC披露声明:如果有联盟/赞助内容但未披露需标记
  • Key Moments富结果时间戳:格式正确(
    M:SS - 标签
    或
    MM:SS 标签
    ), 符合Google搜索轮播展示条件

4. Tags

4. 标签

  • Count: 5-15 (flag <5 or >20)
  • Tag 1 = exact primary keyword
  • Mix: 1-3 broad + 5-8 specific long-tail + 2-3 brand/channel tags
  • Total chars <500
  • Entity overlap: tags should reference the same entities present in description and transcript (consistency strengthens classification)
  • No tag spam (unrelated popular terms)
  • 数量:5-15个(少于5个或多于20个需标记)
  • 第1个标签 = 精确匹配的主关键词
  • 组合结构:1-3个泛关键词 + 5-8个精准长尾关键词 + 2-3个品牌/频道标签
  • 总字符数 <500
  • 实体重叠度:标签应提及与描述和文字稿中相同的实体 (一致性可强化主题分类)
  • 无标签堆砌(不使用不相关的热门词)

5. Thumbnail

5. 缩略图

Delegate to
youtube-seo-thumbnail
for full treatment. Include a quick score here:
  • Resolution 1280x720, <2MB
  • Readable at 120px wide (mobile test)
  • Face + strong emotion present (niche-appropriate)
  • Text ≤4 words, ≥80pt effective
  • SERP differentiation (pattern-break vs top-10)
  • Consistency with channel grid
  • No title-thumbnail redundancy
完整分析交由
youtube-seo-thumbnail
技能处理。此处提供快速评分:
  • 分辨率1280x720,大小<2MB
  • 120px宽度下可清晰辨认(移动端测试)
  • 有人物面部 + 强烈情绪表达(符合细分领域特性)
  • 文字≤4个词,有效字号≥80pt
  • SERP差异化(与前10名结果相比打破固有模式)
  • 与频道整体风格一致
  • 标题与缩略图无信息冗余

6. Chapters & Key Moments

6. 章节与Key Moments

  • ≥3 chapters
  • First is
    0:00
  • Each ≥10s
  • Descriptive, keyword-rich labels (not "Intro / Main / Outro")
  • Matches actual structure
  • Schema check: propose
    VideoObject
    with
    hasPart[].Clip
    and
    SeekToAction
    entries so the embed/page unlocks Google "Key Moments" rich result (see optimize skill for code)
  • 章节数≥3
  • 首个章节为
    0:00
  • 每个章节时长≥10秒
  • 标题具有描述性、包含丰富关键词(不要用“简介/正文/结语”这类通用标题)
  • 与视频实际结构一致
  • Schema检查:建议添加包含
    hasPart[].Clip
    和
    SeekToAction
    条目的
    VideoObject
    结构化数据,使嵌入视频/页面可获得Google“Key Moments” 富结果(代码参见优化技能)

7. Captions & Transcript

7. 字幕与文字稿

  • Manual captions present (flag auto-only)
  • Language tag correct
  • Translated captions for top 3 audience countries (from Studio → Audience)
  • Transcript contains primary keyword in first 60s and final 60s
  • Transcript contains ≥5 Knowledge Graph entities from the topic cluster
  • 存在人工字幕(仅有自动字幕需标记)
  • 语言标签正确
  • 为前3大受众国家提供翻译字幕(数据来源:Studio → 受众)
  • 文字稿在前60秒和最后60秒都包含主关键词
  • 文字稿包含≥5个来自主题集群的Knowledge Graph实体

8. Audio Loudness

8. 音频响度

Run FFmpeg loudnorm analysis. Flag:
  • Integrated loudness outside -14 to -13 LUFS (quiet = retention drag)
  • True peak > -1 dBTP (clipping risk)
  • Loudness range >15 LU (inconsistent — causes volume hunting)
执行FFmpeg loudnorm分析。标记以下问题:
  • 整体响度不在-14到-13 LUFS区间(音量过小会拉低留存)
  • 真实峰值 > -1 dBTP(有削波风险)
  • 响度范围 >15 LU(音量不一致,会导致用户频繁调整音量)

9. End-screens & Cards

9. 结束画面与信息卡

  • End-screen uses all 4 slots (best related, playlist, subscribe, channel)
  • Cards at moments of attention (peaks in retention curve if available, else 60-70% mark)
  • Pinned comment: CTA that drives replies (question prompt)
  • Pinned comment reply rate (from Studio if available)
  • 结束画面用完所有4个插槽(最佳相关视频、播放列表、订阅、频道主页)
  • 信息卡放置在注意力高峰时刻(如有留存曲线则放在峰值处, 否则放在视频60-70%进度位置)
  • 置顶评论:设置可带动回复的CTA(提问式引导)
  • 置顶评论回复率(如有Studio数据则提供)

10. Engagement Signals

10. 互动信号

  • Like-to-view ratio vs channel median
  • Comment velocity (comments / views / day since upload)
  • Creator reply rate in first 24h
  • Shares, saves-to-playlist visible
  • Subscribe-per-1000-views (from Studio)
  • Pinned-comment replies
  • 点赞观看比 与频道中位数对比
  • 评论增速(评论数 / 观看数 / 上传后天数)
  • 前24小时创作者回复率
  • 分享数、加入播放列表数(如可见)
  • 每千次观看订阅数(数据来源:Studio)
  • 置顶评论回复数

11. Technical & Safety

11. 技术与合规

  • Resolution ≥1080p
  • Made-for-Kids flag matches content (flag mismatch)
  • Altered/synthetic content disclosure if AI-generated
  • Copyright claims present? (from Studio)
  • Embedding enabled
  • Comments enabled with moderation
  • Default language + audio language tags set
  • 分辨率≥1080p
  • Made-for-Kids标记与内容匹配(不匹配需标记)
  • 如为AI生成内容,需披露内容为修改/合成内容
  • 是否存在版权声明(数据来源:Studio)
  • 已启用嵌入功能
  • 已启用评论并开启审核
  • 已设置默认语言 + 音频语言标签

12. Shorts (if applicable, <60s vertical)

12. Shorts(如适用,时长<60秒的竖屏视频)

If the video is a Short, apply Shorts-specific rules INSTEAD of the long- form retention analysis:
  • 0-1s hook (swipe-away rate)
  • Loopability (does the end tie back to the start?)
  • Caption overlay for muted viewing
  • Audio: trending library audio or original remixable?
  • Title ≤40 chars, front-loaded emotion
  • #Shorts
    present
  • Length sweet spot: 15-30s for loop rate, 45-60s for watch time
  • CTA to long-form via pinned comment or end card
如果视频是Shorts短视频,则采用Shorts专属规则,替代长视频留存分析:
  • 0-1秒钩子(划走率)
  • 循环性(结尾是否能衔接开头?)
  • 提供字幕叠加层,适配静音观看场景
  • 音频:是热门库音频还是可二次创作的原创音频?
  • 标题≤40字符,情感点前置
  • 包含**
    #Shorts
    **标签
  • 时长黄金区间:15-30秒利于提升循环率,45-60秒利于提升观看时长
  • 通过置顶评论或结束卡片引导观看长视频

Output

输出内容

Video Score Card

视频评分卡

Overall Score: XX/100   (vs niche median: X.Xx)

Retention:         XX/100  ████████░░   [requires Studio CSV]
Title:             XX/100  ██████████
Description:       XX/100  ███████░░░
Tags:              XX/100  █████░░░░░
Thumbnail:         XX/100  ████████░░
Chapters:          XX/100  ██████░░░░
Captions:          XX/100  ███████░░░
Audio Loudness:    XX/100  ████████░░
Engagement:        XX/100  ███████░░░
Technical/Safety:  XX/100  █████████░
Overall Score: XX/100   (vs niche median: X.Xx)

Retention:         XX/100  ████████░░   [requires Studio CSV]
Title:             XX/100  ██████████
Description:       XX/100  ███████░░░
Tags:              XX/100  █████░░░░░
Thumbnail:         XX/100  ████████░░
Chapters:          XX/100  ██████░░░░
Captions:          XX/100  ███████░░░
Audio Loudness:    XX/100  ████████░░
Engagement:        XX/100  ███████░░░
Technical/Safety:  XX/100  █████████░

Root-Cause Ranking

根本原因排序

Rank the 3 largest drags on performance in order, each with estimated impact (e.g., "Hook problem — fixing 0:15 retention from 55% → 70% adds ~20% AVD, estimated +30% impressions").
按影响程度从高到低排序3个最大的性能拖累因素,每个都要附上预估影响 (例如:“钩子问题——将0:15的留存率从55%提升至70%,可增加约20%的AVD, 预估曝光量提升30%”)。

Issues Found

发现的问题

Critical → High → Medium → Low with one-line fixes and Tier reference.
按严重程度分为 严重 → 高 → 中 → 低,每个问题附带一行修复方案和优先级参考。

Paste-Ready Rewrites

即贴即用的改写方案

Hand off to
youtube-seo-optimize
for full rewrites OR inline:
  • 5 title alternatives (tagged Browse-primary / Search-primary)
  • New description body (with chapters, entities, hashtags)
  • Tag list
  • VideoObject
    +
    Clip
    +
    SeekToAction
    JSON-LD schema block
  • Pinned comment CTA
交由
youtube-seo-optimize
技能完成完整改写,或直接提供以下内容:
  • 5个备选标题(标注为Browse优先/搜索优先)
  • 新的描述正文(包含章节、实体、话题标签)
  • 标签列表
  • VideoObject
    +
    Clip
    +
    SeekToAction
    JSON-LD结构化数据块
  • 置顶评论CTA

Measurement Plan

效果衡量计划

ChangeMetric to watchWindowSuccess threshold
New titleBrowse CTR7 days+15% vs baseline
New thumbnailImpressions CTR14 days+1 absolute pt
Fixed intro0:30 retentionnext upload≥70%
Key Moments schemaGoogle Search video clicks30 daysfirst rich result
改动项监控指标观察周期成功阈值
新标题Browse CTR7天较基准提升+15%
新缩略图曝光CTR14天绝对值提升+1个百分点
优化开篇0:30留存率下次上传≥70%
Key Moments结构化数据Google搜索视频点击量30天首次获得富结果展示

Error Handling

错误处理

ScenarioAction
No Studio CSVScore retention as "unknown"; do not fabricate a curve
Tags not visibleUse API or ask; never invent a tag list
Video <48h oldFlag that velocity and CTR data are too early to judge
Private/unlistedAsk for unlisted-shareable link or metadata paste
Shorts videoUse Shorts ruleset; skip long-form retention checks
Live stream VODAnalyze the VOD as long-form but flag that chat spikes distort retention
场景处理方式
无Studio CSV将留存率评分为“未知”;不得编造曲线
标签不可见使用API或询问用户;绝不能编造标签列表
视频发布不足48小时标记增速和CTR数据尚早,无法判断
私密/未公开视频索要可分享的未公开链接或粘贴元数据
Shorts短视频使用Shorts规则集;跳过长视频留存检查
直播回放VOD按长视频分析VOD,但需标记聊天峰值会扭曲留存数据