Transcript Dock

If you have ever wired up caption scraping for one platform, then a speech-to-text fallback, then format converters, Transcript Dock is that pipeline as a single API. Every source goes through POST /v1/jobs and comes back in the same job shape, moving through stages such as resolve, captions and finalize. Each job reports its status, and failures come back as typed errors. A YouTube video with no captions returns NO_CAPTIONS instead of a blank transcript that looks like success.

For YouTube and TikTok links, it fetches the creator's own captions, which costs one credit and is charged only when the job succeeds. When there are no captions, or the source is an upload, a direct HTTPS media link or an Instagram Reel, paid plans run AI transcription instead at two credits per started minute. That path detects the spoken language and returns word-level timing with confidence scores. Stored transcripts can be exported as JSON, SRT, VTT or TXT as often as you like without paying again.

Developers adding transcripts to a product, automation builders feeding video content into workflows, and people who want an AI agent to read videos. Batches accept 10, 25 or 50 links depending on the plan, signed webhooks report when a job finishes, and an MCP server lets Claude, Cursor or VS Code create jobs and read transcripts directly.

One thing to weigh before building on it. Transcript Dock fetches captions and audio from YouTube, TikTok and Instagram on your behalf, and its pricing page lists routing those fetches through proxies, with cooldowns and retries, as part of the service. Pulling content from these platforms this way can conflict with their own terms, so check that it fits how you plan to use the transcripts.

The free plan gives 50 caption credits every 30 days with no card, for YouTube and TikTok only, with seven-day retention. Starter is $19 a month for 2,000 credits, Pro is $49 for 6,000 and Scale is $149 for 20,000, with extra credits from $6 to $8 per thousand, retention of 30 or 90 days, and yearly billing at about two months free. Failed jobs cost nothing.

The privacy policy, updated October 31, 2026, names VOLODS LTD, registered in England and Wales, as the controller. It lists its processors, including Supabase, Vercel, Oracle Cloud, Google Cloud and AssemblyAI for AI transcription, and says audio is deleted from its storage after processing and content isn't used to train models. For videos that mention other people, it acts as your processor and offers a data processing agreement on request.