Google Cloud Translation review

Google Cloud Translation is a managed Google Cloud service for adding machine-learning translation and language detection to apps and workflows. It supports text and document translation across up to 189 languages, according to the published product information. Basic (v2) handles real-time text and HTML, with automatic source-language detection. Advanced (v3) adds document translation, batch jobs, glossaries, and tools for adapting or customizing output. It can translate PDF, DOCX, PPTX, and XLSX files; the product information says formatted documents preserve formatting and images. Adaptive Translation uses example pairs to guide tone and style, while custom models can be trained through AutoML Translation or Vertex AI pipelines. Large jobs can run asynchronously through Google Cloud Storage buckets. Google also offers Translation Hub for document workflows and an isolated on-premises option through Google Distributed Cloud air-gapped appliances.

The range of workflows is the main advantage. Basic covers real-time text and HTML translation; Advanced adds formatted document translation, batch processing, glossaries, romanization, and language detection. That gives teams options beyond pasting text into a consumer translation tool, including ways to keep document layout and images intact.

Customization is useful when generic output is not enough. Glossaries apply set translations to names and specialized terms, while Adaptive Translation can use example pairs to guide style without retraining a model. Teams with deeper requirements can train custom models through AutoML Translation or Vertex AI pipelines.

The pricing page lists a monthly free allowance of 500,000 characters for Basic and Advanced, with a $10 credit shared between them. NMT text translation and several Advanced features cost $20 per million characters above the free tier. There is also a 90-day trial with $300 in credits for new Google Cloud customers. For teams with irregular workloads, pay-as-you-go billing avoids a published subscription commitment.

There is no published monthly or annual subscription price, so costs depend on usage and require estimating your expected volume. The free 500,000-character allowance is shared between Basic and Advanced; it does not cover document translation or LLMs. After the allowance, NMT translation costs $20 per million characters, and usage above 1 billion characters monthly requires contacting sales for custom pricing.

Document costs use different units: formatted DOCX, PPT, and PDF translation is $0.08 per page, or $0.25 per page for custom-model documents. Translation Hub is priced separately at $0.15 per page per target language for Basic and $0.50 for Advanced, with no free plan listed. Custom AutoML text rates range from $80 to $30 per million characters across volume tiers, and training is $45 per hour, capped at $300 per job. A Reddit discussion also notes that the free consumer translate.google.com service is not the same as using the API in another tool; API usage can incur charges.

There’s no monthly or annual subscription price: the pricing page lists pay-as-you-go rates. Basic and Advanced each include the first 500,000 characters per month free, with a $10 credit shared between them; Advanced excludes documents and LLMs from that allowance. Above the free tier, standard NMT translation costs $20 per million characters on either plan. Above 1 billion characters monthly, Google says to contact sales.

Advanced adds costs for specific workflows: formatted DOCX, PPT and PDF translation is $0.08 per page, or $0.25 per page with a custom model. Translation LLM usage is $10 per million input and output characters; Adaptive Translation LLM is $25. AutoML text translation ranges from $80 to $30 per million characters across usage tiers, plus $45 per training hour, capped at $300 per job. Translation Hub is separate: $0.15 per page per target language for Basic and $0.50 for Advanced. New Google Cloud customers can get 90 days and $300 in credits. Track usage: document, model and LLM charges can sit outside the shared character allowance.

The work depends on what you use. A basic text integration needs a translation service in your application; Advanced workflows may also rely on document formats, Cloud Storage batch jobs, glossaries, adaptive examples or custom models. Replacing those means checking which equivalent features your next provider offers and adapting the relevant workflows. That is a likely migration consideration, not a published estimate of effort.

The published material does not specify data-export tools, migration utilities or how custom models transfer to another provider. We’d confirm those details before switching, especially if your process depends on trained models or saved terminology. The clearest reason to leave in our verdict is cost predictability: usage-based charges can be hard to budget, and occasional users may not want a metered API. We can’t substantiate a Reddit-based account of why customers leave from the published sources, so we won’t attribute one. Before moving, inventory API calls, document workflows, glossaries and model assets, then confirm what can be exported and what must be rebuilt.