Short version: if you only need YouTube captions and can run Python on a residential IP, the open source youtube-transcript-api is free and good. If you need TikTok, a hosted endpoint, word timing, subtitle exports or an MCP server for Claude and Cursor, Transcript Dock or Supadata is the shorter path. If you already have audio files and want the best raw recognition, go straight to AssemblyAI.
All ten side by side, with prices per transcript: video transcript APIs compared. Detailed pages: vs Supadata, vs TranscriptAPI, vs youtube-transcript-api and vs Apify scrapers.
Side by side#
| Feature | Transcript Dock | youtube-transcript-api | Supadata | AssemblyAI |
|---|---|---|---|---|
| Sources | YouTube, TikTok and Instagram Reels links; file uploads and direct links | YouTube only | YouTube, TikTok, Instagram, X, Facebook | Audio file or audio URL you supply |
| Existing captions | Creator and platform captions, word timing when the track has it | Yes (creator and auto-generated) | Yes | No, always runs recognition |
| AI transcription | YouTube, TikTok, uploads and direct links, 2 credits per minute | No | Yes, 2 credits per minute | Yes, its core product |
| Hosted / needs infra | Hosted API + MCP server | Python library on your machine; cloud IPs are blocked without a proxy | Hosted API | Hosted API |
| MCP server | Remote endpoint with OAuth sign in, and npx package | No | Remote endpoint and npx package | No first-party server listed |
| Exports | JSON, SRT, VTT, TXT | Objects with start and duration; SRT/WebVTT formatters | JSON | JSON, SRT, VTT |
| Free tier | 50 credits every month, no card | Free (open source) | 100 credits per month | $50 of free credits |
| Entry price | $19 per month, 2,000 credits | $0 (plus a proxy if you run in the cloud) | $5 per month billed annually, 300 credits | $0.15 per audio hour (Universal-2) |
youtube-transcript-api#
A Python library that fetches the transcript YouTube already has for a video, with no API key and no browser. Its README notes that YouTube "has started blocking most IPs that are known to belong to cloud providers", so production use from AWS, GCP or a VPS needs a rotating residential proxy, which the library supports. It does not cover TikTok and does not do AI transcription.
- Pick it when: YouTube only, a script or notebook, you control the IP.
- Pick Transcript Dock when: you need a hosted endpoint, TikTok, batch jobs, webhooks, exports or MCP.
Supadata#
A hosted transcript API covering YouTube, TikTok, Instagram and X, with optional AI transcription for videos that have no captions (2 credits per generated minute per its pricing page) and integrations for Make, Zapier, n8n and an MCP server. It is the closest alternative in shape.
- Pick it when: you need X or Facebook today, or their no-code integrations.
- Pick Transcript Dock when: you want typed error codes with a retryable flag, idempotent submits, a versioned OpenAPI contract, per-job credit caps, and a transcript cache so re-exports never cost a second request.
AssemblyAI#
A speech-to-text API: you give it an audio file or URL and it returns a transcript with word timings. It does not ingest YouTube or TikTok links or use existing captions, so for social video you would still fetch and extract the audio yourself. Its listed price is $0.15 per audio hour for Universal-2 and $0.21 for Universal-3.5 Pro.
- Pick it when: you already have audio and want to own the pipeline.
- Pick Transcript Dock when: the input is a link. Captions are fetched first (no recognition cost) and recognition is only used when a video has none.