Skip to main content
Amazon Translate is AWS’s neural machine translation service, covering 75 languages and variants — including languages that are thinly supported elsewhere, such as Afrikaans, Armenian, Azerbaijani, Dari, Hausa, Haitian Creole, Pashto, Sinhala, Somali, Uzbek, and Welsh. Its distinguishing value in this catalog is output control rather than raw model capability.
Available in Ollang Workflows as aws — Document Translation, Subtitle Translation, and AI Dubbing. See the Translation catalog.

​
Customization controls

Brevity is the underrated one for localization: most engines will happily produce a target string 40% longer than the source, which then has to be cut by hand to fit the subtitle.

​
Key capabilities

  • 75 languages with real-time document translation supported from English into any supported language and back.
  • Real-time and batch processing from the same service.
  • AWS-native security — encryption in transit, with IAM, VPC, and regional controls applying as they do across AWS.
  • Direct chaining with Amazon Transcribe and Amazon Polly, and with S3, Lambda, and Comprehend for pipeline work.

​
Where it fits in a workflow

For organizations with a mature translation memory and hard terminology requirements, the Custom Terminology and Active Custom Translation combination is often worth more than a higher-capability general model that has to be re-steered on every Order. It is also a natural fit where the wider pipeline already runs on AWS, alongside Amazon Transcribe in the transcribe step.

​
Trade-offs

  • NMT, not an LLM. No nuanced instruction-following, cultural adaptation, or reasoning about ambiguous source text.
  • No general quality advantage over DeepL or Agentic Turbo — its case rests entirely on terminology control, Brevity, and AWS alignment.

​
Reference