Skip to content

TikTok Transcript MCP

Remote MCP server that transcribes public TikTok videos by listening to the audio with an AI speech model, not by copying TikTok's auto-captions. Because it runs speech recognition itself, it works on videos with no captions at all and across the 90+ languages TikTok creators actually post in. Per video it returns clean punctuated text, word- or sentence-level timings, speaker labels for duets and interviews, ready-to-use SRT and VTT subtitle files, per-segment confidence scores, and optional post metadata (author, hashtags, views, likes).

Server type: Remote (Streamable HTTP) on the Apify MCP gateway
Auth: Your own Apify token or Apify OAuth (no anonymous mode)
Endpoint: https://mcp.apify.com?tools=deapi/tiktok-transcript-scraper
Tools: 1 actor surface (transcription with rich per-video output)
Pricing: Apify actor runs; private, deleted, and region-locked videos return failed rows and are not charged
Category: Content & Research / Social content intelligence
Built by: deapi-ai (GitHub deapi-ai/tiktok-transcript-mcp, registry io.github.deapi-ai)

Why This Matters for Operators

TikTok is where competitor messaging, creator language, and category narratives move first, but it is a video wall: nothing is greppable, and captions only exist when the creator adds them. Transcription converts that wall into text an assistant can search, summarize, and trend over. A content operator can pull transcripts from a competitor's last twenty videos and ask for recurring hooks, pricing mentions, and objection language; a market researcher can track how a category is being talked about week over week.

The output format matters too. Speaker labels handle duets and interview formats, and the SRT/VTT files mean the transcripts drop straight into editing and subtitling workflows. The failure model is operator-friendly: private, deleted, and region-locked videos come back as failed rows and are not charged.

TikTok content becomes readable, searchable research material.

Tools & Capabilities

Capability-level table from the repository documentation; the endpoint requires Apify authentication (verified: unauthenticated requests are rejected with 401 before the MCP handshake).

Tool Capability
tiktok-transcript-scraper Transcribe public TikTok videos by URL: punctuated text, timings, speaker labels, SRT and VTT subtitles, confidence scores, and optional metadata (author, hashtags, views, likes)

Installation

claude mcp add tiktok-transcript --transport http "https://mcp.apify.com?tools=deapi/tiktok-transcript-scraper"

Authenticate with a token from Apify Console, Settings, API and Integrations, or let the client open Apify's OAuth flow.

Configuration

{
  "mcpServers": {
    "tiktok-transcript": {
      "type": "http",
      "url": "https://mcp.apify.com?tools=deapi/tiktok-transcript-scraper",
      "headers": {
        "Authorization": "Bearer <your-apify-token>"
      }
    }
  }
}

Business Relevance

  • Content operators transcribe competitor feeds to mine hooks, offers, and objection language.
  • Market researchers track how a category is discussed on TikTok across languages and weeks.
  • Brand teams audit creator mentions of their product with metadata like views and hashtags attached.
  • Agencies pull SRT/VTT subtitle files straight into edit workflows for repurposing.
  • SEO and content teams feed transcripts into topic research for short-form and long-form content alike.

Integration with CorpusIQ

The transcript MCP turns TikTok into text; CorpusIQ turns that text into business context. A composed workflow: the assistant transcribes a competitor's recent feed, then joins the trending topics with Shopify sales by product and GA4 landing-page traffic via CorpusIQ connectors to see which narratives actually moved revenue. Social listening and business results in one conversation.

Limitations

  • No anonymous mode: every request needs an Apify token or OAuth session.
  • One actor surface: transcription only, no comments, profiles, or follower-graph data.
  • Apify actor runs are metered; batch pulls accumulate cost (capped budgets recommended).
  • Private, deleted, and region-locked videos fail by design and return failed rows.
  • Young listing: repo created Aug 27, 2026, 0 stars, no license declared in the GitHub metadata.

See Also

Contact Terms Privacy Pricing