Tracking Expenses by Voice: Say It, Confirm It, Done

Tracking Expenses by Voice: Say It, Confirm It, Done

People drop finance apps not because the reports are too few. They drop them because recording every purchase takes too long: open the app, find the account, type the amount, pick a category, fix the date. A week without entries turns into a backlog you owe yourself, and the app gets closed for good.

In ExtMoney a purchase can be recorded by voice or with a photo of the receipt. You say "coffee, two fifty, cash", the app drafts the transaction, you check it and confirm. Below: how it works, what happens to the recording, and why the last word is always yours.

How to record an expense by voice

1. Tap the microphone on the transactions screen.
2. Say what you bought, how much, and how you paid. For example: "groceries at Pyaterochka, one thousand two hundred, card".
3. The app shows a draft: amount, category, account, date. Check it and tap "Save".

A recording lasts up to 20 seconds. You can speak Russian, English, German, Spanish or French. Dates are understood the human way: "yesterday", "on Friday", "on the third".

Speech is recognised by Yandex SpeechKit, and the draft is assembled from the text by YandexGPT. Both run in Russia; the data does not leave the country.

Or take a photo of the receipt

Not every purchase is convenient to say out loud: a stack of receipts after a trip, a paper bill, a noisy place. Point the camera at the receipt. The amount, date and store are recognised on their own and go into the same draft. If the receipt has several items, the app offers to split the purchase across categories.

Voice and photo are two doors into one room. Either way you get a draft, and nothing enters your books until you have looked at it.

Why the last word is yours

Recognition makes mistakes: background noise, an accent, a faded receipt. So ExtMoney does not let the model decide your transaction on its own. Every draft goes through checks written in code, not by a neural network:

• "yesterday" becomes a calendar date, not the model's guess;
• the amount has to be a number within sensible limits;
• the category and the account have to match ones you already have. If there is no match, the app asks instead of inventing a new category.

If everything matched, the draft is ready to save with one tap. If something is unclear — a vague amount, an unfamiliar category — the app shows what exactly and waits for your answer. That is a little slower, but no stray number settles into your history.

What happens to the recording

Voice and a photo of a receipt are sensitive data, so before the first use the app asks for consent and shows where the recording is sent. Consent can be withdrawn in the settings at any time.

The recording itself is not kept after processing; only the text remains. Card and phone numbers are removed from it before it is saved. The receipt photo can be attached to the transaction if you want to keep it as a document.

Worth knowing

Voice input and receipt scanning are available on every plan, with a monthly limit. On the free plan it is 20 voice entries and 5 receipts: enough to try it on your own purchases. On Plus it is 300 and 50, on Family 1,500 and 200.

The app does not connect to your bank. Voice and photo replace typing, not a bank statement: you still record the purchases yourself, it just takes a few seconds.

Getting started

1. Add the accounts you pay from: your card, cash.
2. Record the first purchase by voice and look at the draft. Fix the category if needed: the app remembers it and fills it in next time.
3. Take a photo of the nearest receipt.
4. Record purchases the same day. That keeps the books honest, and no backlog builds up.

You can start for free: ExtMoney.