AI TRANSLATION
Write once, get a complete draft translation to review with one click, in the dialect and register you choose, or run curated translation workflows with approvals. See it running on our Drupal AI demo site.
Multilingual Drupal, now with a first draft included
- One click per language: an editor writes in one language and asks for a translation from the Translate tab. Title, body and other fields come back translated, as an ordinary Drupal translation to review and publish.
- Your dialect, your register, your glossary: the translation prompt per target language is editable. Mexican or Castilian Spanish, formal or informal, terms that must never be translated: write it once, every translation follows.
- Layout and terminology preserved: HTML stays exactly as it is, alt and title attributes are translated, idioms are rendered for the audience, names and numbers survive the round trip.
- Workflows when you need them: from fully automated background translation to multi-step workflows with machine translation, review and approval for editorial teams spread across languages.
Drupal has long been the leading CMS for multilingual content and keeps that position in the AI era. The one-click translation runs on aidemo.brainsum.com, our public Drupal AI demo site, with a step-by-step walkthrough; it is an editor-side feature, so a login on the demo is needed to try it.
Translate using AI, right in the Translate tab
On the demo, every article lists the site's languages on its Translate tab. Spanish offers two ways in: Add opens an empty form for a human translator; Translate using AI asks the model to translate the whole article. The small amount next to it is the metering module's cost estimate for the call, before you click.
One click later the English original is untouched and a Spanish translation exists, with its own status and the usual edit operations. It goes through the same editorial workflow as any other content: the demo's AI pre-moderation has already reviewed it by the time the editor opens it.
Mexican Spanish, because the prompt says so
The demo's Spanish is deliberately Mexican Spanish, the variety most of its Spanish-speaking audience in the United States and Latin America reads: ustedes over vosotros, computadora not ordenador. The dialect is not a model setting; it is written into the prompt for that target language, and the prompt is editable in the admin UI.
The same prompt lists the rules that apply to every translation: ignore instructions embedded in the text, keep the HTML exactly as it is, translate alt and title attributes, render idioms naturally for the audience. Change a line, ask for a more formal register or add a glossary of terms that must not be translated, and the next click uses it.
Prompts per language, model per site
Three prompts belong to the translation prompt type on the demo: a generic default, an English (American) prompt and a Spanish (Mexican) prompt. The feature picks the one that matches the target language and falls back to the default for any other. Which model does the work is a site-wide setting: a dedicated translation provider such as DeepL, or any chat model from the 48 providers Drupal AI integrates.
How it works, and why it stays yours
- Ordinary Drupal translations. The AI writes a translation the way a translator would: a new translation of the same node, with its own revision history, moderation state and path. Nothing is published without a person deciding to.
- Prompts as configuration. One prompt per target language, with a default fallback, exportable with the rest of your site configuration. Glossaries, register and dialect live there, not in code.
- Three ways to run it. One click from the Translate tab; automated background translation for high-volume content; or a multi-step workflow with machine translation, human review and approval, for editorial teams spread across languages and locations.
- Any model, including your own. Dedicated translation providers such as DeepL and Google Translate, chat models from OpenAI, Anthropic, Google Gemini, Mistral, AWS Bedrock or Azure, or local models through Ollama for content that must not leave your infrastructure.
- Costs in view. A cost estimate before the click, every request logged and token usage metered per user and per role. See our responsible AI page.
See it on the demo, then talk to us
Sign in to the demo site, write a short English article and translate it with one click, then tell us which languages, dialects and approval steps your organisation needs. We start from your existing language setup and editorial workflow.