Description
Azure AI Translator (now part of Microsoft Foundry Tools) is Microsoft’s cloud translation API, offering standard text translation, document translation, and custom model training. It bills on a consumption model, charging per million characters translated, with a free F0 tier for testing and low-volume use. Custom Translation lets an organization train a domain-specific model, billed separately by compute hours rather than characters, while hosting the deployed model is billed hourly. Document translation preserves formatting across common file types, and image-based text (JPG, PNG, BMP, WEBP) can be translated directly, billed per thousand images.
Key Features
- Standard Text Translation — pre-trained neural machine translation billed per character.
- Custom Translation — trains a domain-specific model without building infrastructure from scratch.
- Document translation — translates files while preserving the original source formatting.
- Image translation — extracts and translates text directly from JPG, PNG, BMP, and WEBP images.
- Free F0 tier — a limited monthly character allowance for testing before committing to pay-as-you-go.
- Non-regional resource — available regardless of the Azure region where the resource is created.
How It Works
A developer sends text, a document, or an image to Azure AI Translator’s REST API and gets back a translation using Microsoft’s pre-trained neural models, billed per character (or per thousand images for image-based text). For domain-specific accuracy, Custom Translation trains a dedicated model on an organization’s own parallel data; training is billed by compute hours rather than characters, and the resulting model is billed hourly once deployed and hosted. Document translation preserves the source file’s formatting rather than returning plain text, and everything runs on Azure’s standard consumption pricing, with a limited free F0 tier available for testing before scaling to pay-as-you-go or a commitment tier.
Technical Architecture & Overview
- Core Engine: Microsoft’s pre-trained neural machine translation models, plus custom-trained models via Custom Translation.
- Deployment: REST API, part of Microsoft Foundry Tools; container deployment available for high-volume or offline workloads.
- API Surface: Billed per character (standard), per compute hour (custom training), and per thousand images (image translation).
- Known Limits: The free F0 tier doesn’t cover custom model training; commitment tiers and containers are aimed at high-volume enterprise use.
Pros & Cons
| Pros | Cons |
|---|---|
| Standard translation is priced lower per character than several major competing translation APIs. | Custom Translation billing (compute hours plus hosting) is more complex to estimate than simple per-character pricing. |
| Custom Translation lets an organization train a domain-specific model without building infrastructure from scratch. | The free F0 tier doesn’t extend to custom model training. |
| Document translation preserves the original file’s formatting rather than returning plain text. | Getting the most value requires familiarity with the broader Azure ecosystem. |
| Image translation extracts and translates text directly from photos and screenshots. | Commitment tiers and container deployment are aimed at enterprise-scale usage, not casual developers. |
| A $200 free Azure credit for new accounts can be applied toward custom model training costs. | Pricing calculations can vary by region and currency exchange rates. |
Pricing
| Plan | Price |
|---|---|
| Free (F0) | $0 (limited characters/month) |
| Standard Translation | $10 per million characters |
| Custom Translation | Billed by compute hour (training) + hosting |
https://azure.microsoft.com/en-us/pricing/details/translator/
Platform Availability
REST API | Container Deployment
Best For
Developers building multilingual apps already on Azure | Organizations needing a custom-trained translation model for their domain | Businesses translating scanned documents or images
Similar Tools in This Directory
Also worth comparing: Amazon Translate, AWS’s pay-per-character neural translation API | Google Translation AI, Google Cloud’s neural translation API and AutoML custom models | DeepL Translator, a translation tool known for natural-sounding European language output | Phrase Language AI, an enterprise localization platform with AI-driven engine selection
Visit Official Website
Frequently Asked Questions
Does Azure AI Translator have a free tier?
Yes. The F0 free tier gives a limited number of characters per month for standard text translation, according to Microsoft’s official pricing page.
How much does Azure AI Translator cost?
Standard Text Translation costs $10 per million characters, while Custom Translation is billed separately by compute hours for training plus hourly hosting once deployed, according to Microsoft’s official pricing page.
Can Azure AI Translator translate documents?
Yes. Document translation preserves the source file’s original formatting rather than returning plain text, according to Microsoft’s official site.
Does Azure AI Translator support custom models?
Yes. Custom Translation trains a domain-specific model on an organization’s own data, billed by compute hours for training and hourly for hosting, according to Microsoft’s official documentation.
Can Azure AI Translator read text from images?
Yes. It translates text directly from image formats like JPG, PNG, BMP, and WEBP, billed per thousand images, according to Microsoft’s official pricing page.
