Skip to the content
Agent Tools

Transcribe Audio or Video to Text and SRT Captions

Transcript Kit answers four short questions about a recording you own or may use: what is said in it, which parts make good short clips, where a topic comes up, and how the captions read in another language. Last checked 2026-10-24.

Connection
Streamable HTTP
API key
None needed
Rate limit
60 calls a minute
Protocol
MCP 2024-11-05

What does Neil Armstrong say in the first-step recording?

The transcript reads "I'm going to step off the limb now. That's one small step for man. One giant leap for mankind." That is 19 words in 24 seconds of NASA audio. Armstrong said "LM", the lunar module, where the model wrote "limb". The model also did not hear an "a" before "man".

What the live server returned on 2026-10-24
0 to 7 sI'm going to step off the limb now.
7 to 18 sThat's one small step for man.
18 to 24 sOne giant leap for mankind.
Languageen
Speech model@cf/openai/whisper-large-v3-turbo

The call: transcribe_media with {"url":"https://upload.wikimedia.org/wikipedia/commons/d/dd/Armstrong_Small_Step.ogg","confirm_rights":true,"title":"Apollo 11 first step","language":"en","include":["segments","srt"]} on https://transcript.openkrill.app/mcp, run on 2026-10-24 at 07:33 UTC, HTTP 200. The answer above is read from that reply; a later call can return newer data.

Frequently asked questions

Transcribe an audio file to text with timestamps?

transcribe_media turns one audio or video file into timestamped text segments with SRT and VTT captions, and detects the language.

Give an uploaded file or a link to a public social video or a direct media file, up to 15 MB or 15 minutes, in mp3, m4a, mp4, wav, ogg, flac or webm. YouTube links are not taken. A daily allowance of free minutes applies, the transcript is kept for 24 hours and the media file is never kept. Tool: transcribe_media on https://transcript.openkrill.app/mcp.

Find the best 60 second clips in my talk?

suggest_clips returns up to 5 short clips from a saved transcript, each with its start and end, a hook line, the text, a score from 0 to 100 and the reasons.

The clip length defaults to 30 to 60 seconds. It returns text only and does not cut or post video, and the scores rank clips against each other without predicting views. Tool: suggest_clips on https://transcript.openkrill.app/mcp.

Where did I talk about the launch in my recordings?

search_transcripts finds passages in your transcripts from the last day, with the start and end time of each, and lists the themes that repeat.

Only your own latest 10 transcripts are searched, and each is kept for 24 hours. Tool: search_transcripts on https://transcript.openkrill.app/mcp.

Translate SRT subtitles to Spanish?

translate_subtitles translates an SRT or VTT file or a saved transcript, keeps the original times and writes a dub script that marks each line as ok, tight or long.

One call takes about 18,000 characters, about 20 minutes of speech, and nothing is stored. The translation is machine output to check before you publish it. Tool: translate_subtitles on https://transcript.openkrill.app/mcp.

Run it yourself

The answer at the top of this page comes from a real call on 2026-10-24. Call the tool on the server for current data. The server is Transcript Kit at https://transcript.openkrill.app/mcp, with no key. Install steps are on the Transcript Kit page. The full tool list is in the install guides.

Source: the Transcript Kit plugin listing in this repository and a live tools/call on 2026-10-24. Last checked 2026-10-24. These answers are reviewed again every 1 to 3 months.