Як сканувати чеки та автоматично категоризувати витрати
У мене є коробка з-під взуття. Вона стоїть під столом. Усередині приблизно 200 чеків за останні шість місяців, які повільно вицвітають і стають нечитаємими. Якщо це вам знайомо, ви вже розумієте проблему. Чеки накопичуються, бо розбирати їх здається гіршим, ніж просто ігнорувати. До податкового сезону, звісно. Застосунки для сканування чеків якраз і потрібні, щоб це виправити, а хороші не лише читають текст, а й самі розуміють, що ви купили.
- Money Vault сканує чеки прямо на пристрої через Apple Vision. Без завантаження в хмару й без інтернету.
- Автокатегоризація призначає кожному чеку потрібну категорію за назвою магазину та рядками покупки
- Підтримує 50+ валют і міжнародні формати чеків (дати, десяткові розділювачі, формат магазину)
- Чеки вицвіли? Використовуйте спалах, розправте папір і за можливості скануйте протягом 48 годин
У цьому гайді
Як цей гайд робить сканування точним
Потік сканування тут завжди йде в одному порядку: розправити чек, захопити повний кадр, а потім перевірити суму й категорію перед збереженням. Так процес залишається швидким і не пропускає помилки OCR.
- Скануйте, поки чек ще читається, а не через кілька тижнів.
- Спочатку перевірте суму та магазин, а потім, за потреби, рядки.
- Виправте один неправильний чек, щоб розумний кеш став кращим наступного разу.
Чому сканування краще за ручне введення
Manual entry means looking at a receipt, typing the amount, picking a category, adding a date, maybe writing a note. That's 30-45 seconds per receipt if you're quick. Five receipts from a weekend trip? That's nearly four minutes of data entry. Most people just won't do it.
Scanning takes about 5 seconds per receipt. Point your phone, tap capture, confirm the result. The app pulls the total, date, merchant name, and individual line items automatically. You spend your time confirming, not typing.
But the real value of scanning isn't just speed. It's accuracy. When you type "$42.50" from memory, you might actually mean $42.05. Or $43.50. Or that was the subtotal, not the total after tax. The receipt has the actual number. Let the camera read it.
Як сканувати чек покроково
Here's the process in Money Vault. Takes about 10 seconds total.
- Open the app and tap the camera icon. It's on the home screen, next to the voice button. First time, you'll grant camera permission. Standard iOS dialog.
- Position the receipt. Lay it on a flat, dark surface. Hold your phone parallel to the receipt, not at an angle. Get the whole thing in frame. You don't need to crop perfectly. The app finds the edges automatically.
- Tap to capture. One tap. The OCR engine (Apple Vision, running locally on your phone) reads the text in about a second.
- Review the parsed data. The app shows you what it found: total amount, date, merchant name, and a list of individual items if it detected line items. Each field is editable. Tap anything that's wrong and fix it.
- Confirm and save. The expense gets logged with the auto-assigned category and your default account. Done.
Everything happens on your device. The receipt image, the OCR processing, the categorization. Nothing gets uploaded anywhere. If you're scanning a medical receipt or something personal, that matters a lot.
Як працює авто-категоризація
After the OCR extracts text from your receipt, the categorization engine kicks in. It works in two passes.
Pass 1: Merchant matching. The app checks the merchant name against a database of known businesses. "Whole Foods" maps to Groceries. "Shell" maps to Transport. "CVS" maps to Health. This catches about 70% of receipts from recognizable chains and well-known stores.
Pass 2: Keyword analysis. For unknown merchants or generic names like "Corner Store" or "Market #7," the engine looks at the line items. If it sees "milk," "bread," "eggs," that's Groceries. If it sees "ibuprofen" and "bandages," that's Health. This second pass catches most of what merchant matching misses.
When neither pass is confident enough, the app assigns the most likely category and flags it for review. You'll see a small indicator on the expense. Tap to change it, and the smart cache remembers your choice for next time.
Correct miscategorized receipts in your first week. The smart cache uses 85% similarity matching, so fixing "Corner Market" once means similar transactions ("Corner Market #2," "The Corner Market") also get the right category going forward.
Як працювати з багатопозиційними чеками
A grocery receipt with 25 items is trickier than a coffee shop receipt with one line. Here's how Money Vault handles it.
The OCR reads every line item individually. Each one gets extracted with its own amount. You can do three things with them:
- Log the total only. Most common approach. The receipt total becomes one expense entry. Simple and fast.
- Split by category. If you bought groceries and cleaning supplies at the same store, you can assign different items to different categories. Tap individual items and reassign them.
- Log individual items. For detailed tracking, each line item can become its own expense. Overkill for most people, but useful if you're tracking food budgets down to the item level.
Long receipts (the Costco 40-item kind) take a couple extra seconds to process. The parser handles them, though items near the bottom of very long receipts sometimes get cut off if the scan didn't capture the full length. For those, take the photo from a bit farther back so the whole receipt fits in one shot.
Скануйте чеки та забудьте про ручне введення
Money Vault читає чеки й розподіляє їх автоматично. Безкоштовно на iOS.
Що робити з вицвілими чеками
Thermal paper (that shiny stuff most stores print on) starts fading the day it's printed. Heat speeds it up. Body heat from your wallet, a sunny dashboard, even a warm pocket will do it. I've seen receipts become completely blank in three weeks sitting in a hot car.
For faded receipts, try these before giving up:
- Turn on the flash manually. Even in a bright room, the LED flash adds contrast that helps the OCR pick up faint text. It recovers more than you'd expect.
- Try different angles. Thermal paper sometimes reflects better at slight angles. Tilt the receipt 10-15 degrees and try again.
- Dark background matters more here. A faded white receipt on a white table is nearly invisible to OCR. Put it on something dark, like a black notebook or a dark desk.
The honest truth: if a receipt is more than 4-5 months old and has been stored in heat, even the best scanner will struggle. The text physically isn't there anymore. Scan receipts within 48 hours of purchase. That's the single best thing you can do.
Чеки з інших країн
Money Vault supports 50+ currencies and auto-detects the currency from the receipt itself. A few things to know about how this works:
Date formats. US receipts show MM/DD/YYYY. Most of Europe uses DD/MM/YYYY. Japan uses YYYY/MM/DD. The parser checks your device locale and the currency context to figure out which format the receipt is using. April 3 won't get confused with March 4 if the receipt is in euros.
Decimal separators. In the US, "42.50" means forty-two dollars and fifty cents. In Germany, "42,50" means the same thing. The OCR engine reads both formats correctly based on the detected currency.
Non-Latin characters. Receipts in Japanese, Korean, Chinese, Arabic, and other scripts get OCR treatment too. Merchant name extraction is less reliable for non-Latin text (Apple Vision performs best with Latin characters), but amounts and dates still parse accurately.
Exchange rates. When you scan a receipt in a foreign currency, the app logs the original amount and converts it to your base currency at the current exchange rate. You can see both amounts in the transaction details later.
7 порад для кращих сканів
- Flatten the receipt. Wrinkles create shadows that the OCR reads as characters. Two seconds of pressing it flat against a table saves ten seconds of corrections.
- Good lighting beats a good camera. A well-lit receipt on a 5-year-old iPhone scans better than a dimly-lit one on the latest Pro Max. Natural daylight is ideal. At night, turn on overhead lights and keep shadows off the text.
- Dark surface underneath. White receipt on white table? The app can't find the edges. Use a black notebook, dark desk, or even a dark phone case laid flat. Contrast helps the edge detection.
- One shot, whole receipt. Don't scan a long receipt in two parts. Stitching introduces errors where the images overlap. Step back and get it all in one frame. Phone cameras have enough resolution to crop later without hurting OCR quality.
- Scan within 48 hours. Thermal paper starts degrading the moment it's printed. The sooner you scan, the cleaner the text. Make it a habit: pay, scan, pocket.
- Check the total, always. Even with high accuracy, roughly 1 in 20 scans will have something slightly off. Glance at the total before confirming. Two seconds of checking prevents a month of accumulated errors.
- Use flash on faded paper. Manual flash in a bright room adds contrast that catches fading text. It's the single most effective trick for older receipts.
Типові проблеми та як їх виправити
Issue: Wrong total extracted. The receipt shows $42.50 but the app parsed $4.25 or $425.00. This usually happens when the decimal point is faded or the total line isn't clearly separated from other numbers. Fix: edit the amount inline. The smart cache will learn from the correction.
Issue: Category seems random. The merchant name wasn't recognized and the line items didn't give enough context. Fix: change the category manually. Do this a few times for the same merchant and the cache picks it up.
Issue: Date is wrong. Often a format confusion (03/04 read as March 4 vs April 3). Fix: tap the date field and correct it. The app adjusts its format detection based on your corrections and locale.
Issue: Long receipt cut off. The bottom items are missing because the scan didn't capture the full length. Fix: retake the photo from farther away. Or just edit the total manually and skip line-item tracking for that receipt.