domains / research / transcript-analysis / transcript-intelligence

transcript-intelligence

Skill card

combined

Both copies describe the same task (turning social video transcripts into hooks, claims, objections, CTAs, and reusable atoms), but each contains material steps the other lacks. The gooseworks copy adds explicit guardrails (never invent spoken content, flag regulated claims, treat engagement as distribution not proof) and a per-video table output with confidence. The ScrapeCreators copy adds concrete API endpoints per platform, a required SCRAPECREATORS_API_KEY, a discovery step for creator/channel inputs, a detailed output template (TL;DR, swipe file, repurposing ideas), and a common-pitfalls section. Crowning one would hide the other's real operational details.

Repositories

Other copies

Combined skill

---
name: transcript-intelligence
description: Turn TikTok, Instagram, YouTube, Facebook, X, LinkedIn, Reddit, or Rumble transcripts into timestamped hooks, claims, objections, proof, calls to action, sponsorship signals, and reusable content atoms. Use directly for transcript analysis or as support for creator, competitor, trend, demand, and repurposing work.
---

# Transcript Intelligence

Analyze what a video says, not just its caption or engagement count.

## Inputs

- One or more TikTok, Instagram, YouTube, Facebook, X, LinkedIn, Reddit, or Rumble post or video URLs.
- The brand, product, or research question.
- Optional comparison dimensions such as hooks, objections, proof, or creator delivery.
- If a creator or channel is provided instead of URLs, first fetch recent posts/videos, then choose relevant videos.

## Transcript Sources

| Platform | Endpoint |
|---|---|
| TikTok | `/v1/tiktok/video/transcript` |
| Instagram | `/v2/instagram/media/transcript` |
| YouTube | `/v1/youtube/video/transcript`, `/v1/youtube/video` |
| Facebook | `/v1/facebook/post/transcript` |
| X/Twitter | `/v1/twitter/tweet/transcript` |
| LinkedIn | `/v1/linkedin/post/transcript` |
| Rumble | `/v1/rumble/video/transcript` |
| Reddit video | `/v1/reddit/post/transcript` |

If a detail endpoint already includes transcript text, use it. If a transcript is unavailable, explicitly mark the item as caption-only; never invent spoken content.

## Workflow

1. Fetch the post metadata with `scrapecreators-api`.
2. Fetch the platform transcript when available. For YouTube, collect public sponsorship signals when relevant. If a transcript is unavailable, explicitly mark the item as caption-only; never invent spoken content.
3. Preserve timestamps and speaker changes. Keep source URL with each transcript.
4. Segment the transcript into: hook/opening, setup/context, main claim or lesson, evidence/examples, payoff, CTA.
5. Extract:
   - opening hook and first payoff;
   - problem, desired outcome, and audience language;
   - product claims and the evidence offered;
   - objections raised or answered;
   - calls to action and disclosed sponsorships;
   - quotable phrases, limited to short excerpts;
   - stories, frameworks, emotional language, and content atoms that stand alone.
6. Compare repeated patterns across videos. Distinguish a single creator's style from category-wide evidence. Cluster by topic and angle. Count recurring themes. Identify repeated hook formulas. Flag the strongest examples with citations.

## Output

Return a markdown report with the following sections:

```markdown
# Transcript Intelligence Report

## TL;DR
- Main themes:
- Strongest hooks:
- Best reusable ideas:

## Transcript-by-Transcript Notes
### [Video title or URL](url)
- Hook: "..."
- Core idea: ...
- Best quote: "..."
- CTA: ...
- Content atoms:
  1. ...

## Patterns Across the Set
| Pattern | Evidence | Example URLs |
|---|---|---|

## Hooks Swipe File
- "..."
- "..."

## Repurposing Ideas
- LinkedIn post:
- X thread:
- Short-form script:
- Newsletter section:
```

Each per-video entry should include source URL, platform, timestamps, hook, claims, proof, objections, CTA, sponsorship signal when available, content atoms, and confidence.

## Guardrails

- Never present an auto-transcript as perfectly accurate. If wording seems garbled, mark it as approximate.
- Flag unclear brand names, numbers, ingredients, prices, and regulated claims for verification.
- Treat high engagement as a distribution signal, not proof that a claim is true or a creative caused the result.
- Do not summarize from the title alone if the user asked for transcript analysis. Fetch transcripts first.
- Do not lose exact wording. Hooks and quotes are more valuable verbatim.
- Do not mix source attribution. Every quote should trace back to a URL.

## Conflicts

- gooseworks-ai/goose-skills specifies a per-video table output with confidence scores; ScrapeCreators/social-media-research-skills specifies a markdown report with TL;DR, swipe file, and repurposing ideas. The combined instruction includes both: per-video entries with confidence within the markdown report structure.