Viral Chaser

Workflow

Step 1 — Check login (skip for public Bilibili videos)

Run: login-manager check <platform>

• platform: douyin | bilibili | kuaishou
• If the result is SESSION_EXPIRED (exit code 2), run login-manager login <platform> first

Step 2 — Run the analyzer

node {baseDir}/scripts/viral_chaser.ts <url> [--no-frames]

• <url>: Full or short-link URL of the video
• --no-frames: Skip key frame extraction (faster, audio-only analysis)

The script outputs a JSON object to stdout. Read it and proceed with analysis.

Output JSON structure:

{
  "ok": true,
  "platform": "douyin",
  "metadata": {
    "contentId": "...",
    "title": "...",
    "desc": "...",
    "author": "...",
    "durationSeconds": 89,
    "coverUrl": "...",
    "stats": { "playCount": 0, "likeCount": 0, "commentCount": 0 }
  },
  "transcript": {
    "text": "全文转录...",
    "segments": [{ "start": 0.0, "end": 5.2, "text": "开场文案" }]
  },
  "frames": ["/tmp/viral_chaser/{id}/frames/frame_00_0s.jpg", "..."],
  "localPaths": {
    "video": "/tmp/viral_chaser/{id}/video.mp4",
    "audio": "/tmp/viral_chaser/{id}/audio.wav",
    "tmpDir": "/tmp/viral_chaser/{id}"
  }
}

Exit codes:

• 0 = Success
• 1 = Error (URL invalid, download failed, etc.) — report to user
• 2 = Cookie expired — run login-manager login <platform>, then retry once

Step 3 — Read key frames (if available)

For each path in frames, use the Read tool to load the image and analyze it visually.

Read: /tmp/viral_chaser/{id}/frames/frame_00_0s.jpg
Read: /tmp/viral_chaser/{id}/frames/frame_01_3s.jpg
...

Analysis Framework

After receiving the JSON output and reading the frames, generate a 追爆报告 in Markdown.

1. 内容摘要

1–2 sentences: what core value does this video deliver to viewers?

2. 开头钩子分析（前 0–10 秒）

Based on transcript.segments where start < 10:

• 钩子类型: 提问型 / 冲突型 / 反转型 / 数字型 / 悬念型 / 痛点型 / 利益型
• 具体文案: quote the exact opening line(s)
• 效果评估: why this hook works (or doesn't)

3. 内容结构拆解

Based on transcript segments, divide into logical sections:

4. 爆款元素评估

Rate each element as 强 / 中 / 弱 with a one-line explanation:

5. 视觉风格分析（基于关键帧图片）

After reading the frame images:

• 色调风格: 暖色系/冷色系/高饱和/低饱和/黑白
• 构图类型: 人脸近景 / 产品展示 / 场景空镜 / 文字卡片 / 混合
• 字幕/文字覆盖: 字体粗细、位置、是否有背景框、动画感
• 整体视觉标签: 3–5 个关键词（如：「真实感」「强对比」「高信息密度」）

If --no-frames was used or frames is empty, note: "（跳过视觉分析，请重新运行不带 --no-frames 参数）"

6. 可借鉴点

3–5 concise, directly actionable techniques. One sentence each.

7. 目标受众

One sentence describing the primary audience persona.

Script Generation

After completing the 追爆报告, ask the user:

"追爆报告已完成。接下来，我可以基于此视频的公式为您生成新脚本大纲。请告诉我：

1. 新视频的主题/产品/角度（或直接回复"按原主题"）
2. 风格偏好：仿写（保留框架换内容）/ 改写（重组表达）/ 创新（仅借鉴钩子类型）"

Then generate the script outline in this format:

【追爆脚本大纲】
来源参考：{platform} - {title}

▍标题备选（5 个，风格各异）
1. ...
2. ...
3. ...
4. ...
5. ...

▍开场钩子（0–8s）
文案：...
画面建议：...

▍主体段一（8–40s）
文案：...
画面素材关键词（用于 pexels-footage 搜索）：...

▍主体段二（40–70s）
文案：...
画面素材关键词：...

▍收尾/CTA（70–90s）
文案：...
画面建议：...

▍制作建议
BGM 风格：...
字幕风格：...
视觉色调：...

After generating the outline, ask:

"脚本大纲已就绪。要直接用 shorts-compose 制作，还是先调整脚本？"

If the user confirms, pass the script to shorts-compose --script "<outline>" or t2video --script "<outline>".

Notes

• Temporary files are stored in /tmp/viral_chaser/{contentId}/ — they are NOT automatically cleaned up so you can re-read frames without re-downloading. Clean up manually with rm -rf /tmp/viral_chaser/ when done.
• Bilibili DASH format: if mediaFormat is DASH, the video and audio streams are separate. The downloaded video.mp4 contains the video stream only; audio is in audio.wav after extraction. This is transparent to the analysis workflow.
• Cookie refresh: if the script exits with code 2 mid-workflow, run login-manager login <platform> and retry. Do not retry more than once.