This is a minor feature release of Translation Manager for Umbraco v17, it conains fixes and new features for making translations quicker and cheap(er).

Batch Support

The big feature for this release is batch support in the machine and AI translation connectors.

Traditionally the Machine connectors (e.g. Microsoft, google, Deepl, and AI models). translate inline as you submit the job, they work through each bit of translatable text and send it to the service to be translated . these models to groupd things together to reduce the number of API calls but its still done in real-time with the site waiting for each translation to be returned before more are sent.

A lot of the machine translation services do support batch submission, that is you can submit a file with all your translations in them and then check back later for when they are done (very much like a traditional translation service).

This has some advantages, - on is its much quicker to send the translations, so there is less waiting around for the process to complete, another is cost, in that batch translations can be much cheaper on some services (e.g. Batch processing is around 50% cheaper for OpenAI Requests).

Improved Translation Memory Support

Another thing we have been working on is improving the local translation memory support for connectors, (this release DeepL now also supports Translation Memory)

Local translation memory is a memory of translations stored locally in your Umbraco site, so content that has already been translated isn't sent again if we know what it is going to be translated to locally.

This saves money and time, it reduces the number or characters or tokens needed to translate a page or collection of pages, as only the changes are sent to translation. thus reducing the amount of work the translation service needs to do, and the amount of tokens you need to spend for it to do it.

[!TIP] Fun Fact: Did you know it's approximately 10 times cheaper to perform AI translations with Translation manager than with Direct AI methods in Umbraco? (this is because translation manager presents just the text to be translated to the AI, which doesn't have to do all the working out of how to get it in and out of Umbraco every time)

and that doesn't take into account how much Translation memory can save you.

Updated Connectors

Updates to connectors shipped with this release include:

DeepL

• The DeepL connector has been updated to support our core Translation Memory and Batch services,

AI Connector

• The AI Connector now supports batch submissions for OpenAI, Azure Models and Microsoft Foundry Translation Models.
• Spit Connector model, you can now see all the individual translation connectors inside the connector menu, meaning you can have even more AI configs if you want to translate different bits different ways

What's new

Fetures

• ⭐ Support for batch submissions at the machine translation level.
• ⭐ Improved Translation Memory support for xliff style connectors
  • ⭐ DeepL Batch Support (submit large jobs to DeepL in one go).
  • ⭐ AI Batch Support for OpenAI and Microsoft Foundry Models.

Fixes

• 🐛 Fix: Quick Translate Button - displays on non-quick enabled sets.

Installing

dotnet package update Jumoo.TranslationManager