# Amazon Textract

6 Hacker News commenters recommended Amazon Textract and 6 criticized it, November 2025 to November 2026. It was recommended in 5 of 13 months.

**Verdict:** Recommended for accurate handwriting recognition and a well-documented Python library; critics call it expensive, prone to surprising costs, and often outperformed by LLMs.

## Why people recommend it

- Good handwriting recognition (1 comments): "it is really good at correctly recognizing hand written characters" ([source](https://news.ycombinator.com/item?id=46168501))
- Well-documented Python library (1 comments): "great python library with the kind of documentation I expect from a python project" ([source](https://news.ycombinator.com/item?id=49478850))

## What people warn about

- Expensive with surprising costs (2 comments): "They're both very expensive and prone to surprising you." ([source](https://news.ycombinator.com/item?id=48645178)) "those top of the line traditional OCR systems like AWS, Textract are themselves extremely slow and computationally expensive" ([source](https://news.ycombinator.com/item?id=46395439))
- LLMs seen as cheaper alternative (2 comments): "I don't think I'd do the Textract step since LLMs have gotten way better and cheaper." ([source](https://news.ycombinator.com/item?id=45798084)) "I can replace the amazon textract requests by a llama.cpp model call" ([source](https://news.ycombinator.com/item?id=47900421))
- Struggles with dense or odd documents (2 comments): "Textract did poorly because it was just too dense for their model" ([source](https://news.ycombinator.com/item?id=49299400)) "a document format that breaks every tool you throw at it" ([source](https://news.ycombinator.com/item?id=47533202))

## Alternatives hackers recommend

- [Google Cloud Document AI](https://hackerapproved.com/p/document-ai.md): 6 recommend

## Where to get it

- [Visit the official site](https://aws.amazon.com/textract/)

## Cite as

Hacker Approved, "Amazon Textract", https://hackerapproved.com/p/amazon-textract/ (counts from 1,795,640 Hacker News comments, November 2025 to November 2026).