Summarize an audio recording
Summarize an audio recording: upload it, get the transcript, and let our own summarizer pick three sentences and up to five key points out of what was said.
Upload the recording and the transcript comes first; the summary is built from that text, not from the sound. Our own summarizer keeps three sentences and up to five key points, and a second tool lists the tasks. Free $0, three files a day, 30 minutes each.
How to get a summary of a recording
- 1
Upload the recording
The widget takes audio and video. A file is up to 30 minutes at the most; the free plan allows three files a day, 30 minutes each.
- 2
Read the transcript once
The summary can only be as good as the text under it. Look through the lines, correct a name or a number in the app, and save.
- 3
Ask for the summary
Ask the plugin for Claude or ChatGPT, or call the MCP server yourself: the tool returns the retelling and the key points in one answer. The plugin and the MCP server are part of Pro.
- 4
Take the tasks as well
A second tool goes over the same text, picks the sentences that name a task, its owner and its deadline when they are there, and returns them as a list.
What you get
- TXT
Plain text for an email, a note or a chat
- SRT
Subtitles for video editors and platforms
- VTT
Subtitles for a browser and a website
- CSV
A table: time, speaker, text
- DOCX
A Word document with speakers and times
- Markdown
Text with headings for notes
What the summary is made of
- The retelling
three sentences — the ones that carry the most repeated content words
- The key points
up to five lines, cut out of the same sentences
- The tasks
the sentences that name a task, with its owner and its deadline when they are there
- Where it runs
on our own server, on the transcript text; no other service is called for it
- Who asks for it
the plugin for Claude or ChatGPT, or the MCP server — both are part of Pro
The summary is built from the transcript
A summary is a reading of the words, and the reading has to be cheap and repeatable. Our summarizer scores every sentence of the transcript by the content words it repeats, keeps the three that score highest in the order they were spoken, and cuts five key points out of the same material. The summary is computed on our own server, from that text, with no call to any other service.
Tips
Fix the transcript before you ask for the summary.
Three sentences are the whole retelling, whatever the length of the recording.
Keep the recording — the summary is only as good as the text.
Ask for the task list as well: it is a second tool over the same text.
Questions this page answers
What exactly does the summary contain?
A retelling of three sentences, in the order they were spoken, and up to five key points cut out of the same sentences.
Is the summary written by a person?
No, and it is not a language model either. It is a rule of our own: the sentences with the most repeated content words win, and the order stays as spoken.
Do I get a list of tasks as well?
Yes. Sentences that name a task, with its owner and its deadline when they are there, are returned as a separate list.
What happens to my file?
A Russian or mixed recording is processed by our own engine, on infrastructure outside the United States. A pure English file may be sent to a third-party cloud speech provider in the United States, when that fallback is enabled on the server.
Reading: 6 minutesUpdated Transcribsi editorial team
Drop a recording — the text arrives in a minute
Speakers by name, a time on every line, export to TXT, SRT, VTT, CSV and Markdown.