Say "twelve euros lunch with Ana" and SpendTrack records it with the amount, date, and type. The voice is transcribed on your own device: only the text travels.
Recording an expense should take as long as it takes to say it.
In SpendTrack, you open the dictation panel, speak, and see the phrase form on the screen as you do it. The transcription is done by your browser using the Web Speech API, on your own phone and in the language you use the app: the audio is not recorded, not uploaded to any server, and not stored anywhere. The only thing that leaves the device is the already written phrase.
This phrase is interpreted by the app's parser, which extracts the amount, currency, merchant, a short concept, type, and date — it understands "yesterday", "on Monday", or "in July" and converts them into a specific day — as well as tags and payment method if you mention them: cash, card, transfer, bizum, or direct debit. Based on what it understands, it applies your rules and what it has already learned from that merchant, so the type is usually pre-set.
Before saving anything, it shows you an "Understood" screen with the large amount, type, date, and merchant. From there, you save with a tap or press Correct, and the expense opens in the full form with everything pre-filled. If it hasn't understood the amount, it tells you and asks you to edit it, instead of making up a figure.
And if the browser doesn't allow dictation, or the mic fails, you won't get stuck: the same screen turns into a text field to write the phrase, which goes through the same parser with identical results.
The audio never leaves your phone
the transcription happens in the browser using the Web Speech API. Only the text you dictated reaches our servers.
No background microphone
SpendTrack only listens while you have the dictation screen open, and stops listening as soon as you finish or close it.
A normal phrase is enough
"twelve euros lunch with Ana", "yesterday 45 at the supermarket with card". No need to dictate fields or follow an order.
Relative dates resolved
"yesterday", "on Monday" or "in July" turn into the actual date of the expense, without you having to look it up in the calendar.
Proposed typology
it is classified among the 14 typologies of the app or your own, applying your fixed rules and what SpendTrack has already learned from that merchant.
You confirm, always
the "Understood" screen shows you what it has extracted before saving it, and Correct opens the full form with everything pre-filled.
If it can't be dictated, it can be typed
unsupported browser or microphone issues, and the same screen switches to keyboard with the same analyzer behind it.
Doubtful amount, clear warning
if you don't understand the figure, it tells you and blocks saving until you review it, instead of slipping in a made-up number.

Create your free account on SpendTrack —the personal expense tracking app from the AIsToHire platform—, open dictation, and speak aloud what you just paid.