Amazon Translate

Amazon Translate Overview

Amazon Translate is AWS's neural machine translation service, built for high-volume, low-cost integration into cloud applications — with custom terminology and active custom learning.

How Amazon Translate Works

Amazon Translate runs inside AWS: call it via API, SDK, or services like S3 batch translation. Custom terminology enforces your glossary at the engine level; Active Custom Translation (ACT) trains a variant on your existing parallel data for domain-specific quality. Batch jobs translate entire document sets in S3 without writing code.

Key Capabilities

• Deep AWS integration (S3 batch translation, Lambda, Comprehend pipelines)
• Custom terminology at the engine level
• Active Custom Translation: train on your parallel data
• 75 languages; per-character pricing among the cheapest at scale
• Real-time and asynchronous batch translation modes

Best Use Cases

• High-volume user-generated content (reviews, chats, listings)
• AWS-native applications needing MT without a new vendor
• Batch translation of document repositories in S3
• Cost-sensitive bulk translation with terminology control

Amazon Translate Pricing

Pay-as-you-go: $15 per million characters for real-time and batch translation, with a 12-month free tier of 2M characters/month. Custom terminology is free; ACT training and usage costs extra. Volume discounts available via AWS commitments.

Frequently Asked Questions

How much does Amazon Translate cost?

$15 per million characters for both real-time and batch, with 2M free characters monthly for the first 12 months. Custom terminology is free; Active Custom Translation has separate training/inference costs.

Can Amazon Translate use my glossary?

Yes — custom terminology applies your approved translations at the engine level, overriding normal output for those terms. Free to upload and use, up to hundreds of terms per list.

Amazon Translate vs Google Translate API?

Similar quality class and pricing ($15 vs $20 per million characters). Amazon wins for AWS-native stacks and batch S3 jobs; Google wins on language count and low-resource pairs. Benchmark your specific pairs.

Verdict

Amazon Translate is the volume-and-integration play: at $15/M characters with S3 batch jobs, it undercuts nearly everything for large corpora, and if you are already on AWS, it is a checkbox rather than a vendor decision. Output quality trails DeepL and frontier LLMs for European pairs — budget for post-editing where quality matters.