Translation with Google Gemini

Already using Gemini for translation? Crowdin turns it into a repeatable localization process. Gemini models work directly on your Crowdin projects and take your glossary, translation memory, style guide and file context into account, so brand terms and tone carry over to every language. Gemini is built into Crowdin, with no app to install, on both crowdin.com and Crowdin Enterprise.

Where Gemini fits in your workflow

• Automatic translation of new content – as soon as source strings arrive, Gemini pre-translates them using your linguistic assets.
• Multi-step AI Pipeline – split the work into stages instead of relying on a single request: context from screenshots and files, the Ambiguity Filter, translation, then file consistency and QA checks. Every stage can run on its own model with its own instructions and verifies the output of the stage before it. Pair a lightweight Gemini model for routine checks with a stronger one for the translation itself.
• Re-translation after improvements – updated your glossary, added context or switched to a newer Gemini model? Run AI translation again on earlier AI output, and anything a person has approved stays as it is.
• Translation through the API – send strings or files to Crowdin's AI translation API and get Gemini translations that follow your TMs, glossaries and style guides, with no project setup required.

Two ways to connect Gemini

• Let Crowdin manage it – Gemini via Vertex AI is ready to use with no configuration. Usage is billed through Crowdin Credits, so you pay as you go.
• Use your Google credentials – connect your Vertex AI or Google AI Studio account to keep billing with Google and manage usage on your side. A natural choice if your company already works in Google Cloud.

Set up Gemini in Crowdin

Open your Crowdin AI settings, enable Google Gemini as a managed service or with your own credentials, and choose it for pre-translation or any AI Pipeline step. The Crowdin AI guide walks through each step.

Screenshot