AI Video Summary

Proof & evidence

Pull spoken prices, dates, and quantities out of a call

1 July 2026

Spoken prices and dates pulled from a call into a list, each tied to the moment it was said.

Prices, dates, quantities, and account-like figures get lost when they only exist in speech. You remember “about forty.” The invoice says something else.

A list, not a spreadsheet fantasy

Spoken numbers extract pulls every clear number with the sentence around it and the time it was said. You can download a short clip for a figure you need to check.

Example: a vendor call becomes a list of amounts and dates you can put next to the invoice.

It does not build a full accounting system. It does not guess a currency that was never named. “A couple of weeks” is not a date.

How to read the list

Start with the figures you already distrust. Play the clip. If the transcript heard “fifteen” and the speaker said “fifty,” believe the file, then fix the row by hand. Speech-to-text is weakest on similar-sounding numbers.

Pair this with Promise ledger when the number is attached to a commitment (“we will invoice 12,000 on the first”). Pair it with Two-recording comparison when last week’s figure moved.

Put the list next to the document

Open the invoice, the quote, or the email thread beside the extract. Tick each spoken figure against the written one. The useful outcome is a mismatch you can play, not a prettier table.

Buyers, journalists, and anyone walking a warehouse with a phone get the same job: do not trust “about forty” when the file still exists.

When the list is empty

If no clear number was spoken, you should not get a decorative table. Go back to the transcript and see whether the figure was only described in words — “the usual rate,” “same as last quarter.” Those are not extractable numbers. Write them as words, or ask the question again on the next call.

Related workflow: Spoken numbers extract

Extract spoken prices, dates, and numbers from a recording | AI Video Summary