title, type, tags, sources, last_updated
| title |
type |
tags |
sources |
last_updated |
| Transcript-Based Summarization |
concept |
| Transcript |
| Summarization |
| YouTube |
| Content-Processing |
| AI |
|
|
2026-04-22 |
Definition
Transcript-Based Summarization 是指从视频/音频内容中提取字幕/ transcript，然后通过 AI 压缩为结构化要点摘要的处理流程。它使长视频/播客的消费从"没时间看完"变为"5 分钟掌握精华"。
Workflow
- Transcript Extraction: 通过 API（TranscriptAPI.com）或 CLI 工具（yt-dlp）获取字幕
- AI Summarization: LLM 处理字幕文本，输出关键点、亮点引用、时间戳
- Structured Output: 生成 bullet points、key quotes、timestamps 等结构化格式
- Delivery: 整合到 Daily-Digest 或 second-brain
TranscriptAPI vs yt-dlp
| Criteria |
yt-dlp |
TranscriptAPI.com |
| Output format |
Verbose CLI logs |
Clean JSON |
| Cloud compatibility |
Fails on GCP/cloud |
✅ Works everywhere |
| Caching |
None |
✅ Cached results |
| Rate limiting |
Random blocks |
✅ Reliable, millions served |
| Dependencies |
Binary required |
HTTP API only |
Applications
Connections
