DeepL Pro & API vs Google Cloud Translation

DeepL Pro & API

Every language my shop needs — German, English, French, Italian, Dutch, Spanish — is on the published list of 126 languages, with regional variants for French, German and Spanish. Glossaries are genuinely useful for product names: grammatical adaptation in context, priority ordering when several are active, terms marked not to be translated, plus formality settings, style rules and custom instructions, and document translation keeps the layout across Word, PowerPoint, Excel, PDF, HTML, InDesign and image files, with an account-scoped adaptive feature that never feeds other customers. It sits below the top because the quality claim rests on blind tests without a published method, and I found no public information on per-segment quality scoring or routing of weak segments to human review.

Google Cloud Translation

Adaptive translation that works from my own small datasets without training, an LLM model alongside the NMT one, glossaries in TSV, CSV or TMX with case handling, and document translation that keeps the layout for Word, PowerPoint and PDF — that covers my six shop languages easily, with 189 languages on the published list. What I did not find is a formality or tone setting, published quality evaluations, or quality scoring outside the enterprise Translation Hub tier.

DeepL Pro & API

Translation memories can be uploaded as TMX, shared with the team, activated in the translator and reused so the same product description is not translated twice, glossaries are managed centrally and shared across teams, and clicking a word in the web translator to pick a better product name is exactly the manual fix I need. However, I found no public information on review or approval stages, automated checks for placeholders or terminology, comments and screenshots for context, or translators restricted to specific languages; the described workflow capability is stated only in general terms. Workable for a shop without a translator on staff, but thin as a managed process.

Google Cloud Translation

Glossaries can be created, updated and deleted entry by entry, which is exactly how I would force the right product name, and the Translation Hub paid tier lists translation memory, human review and quality prediction scores. But nothing on the captured pages describes review stages, automated checks, an in-context editor or translator roles, and glossary version control is explicitly not supported.

DeepL Pro & API

The REST API is documented in real detail — including translation-memory import and export endpoints with their limits and error codes — and there are named connectors for Microsoft 365, Google Workspace, Zendesk, Adobe Experience Manager, Google Drive and SharePoint, plus browser extensions and an AWS Marketplace listing. Files are supported across office formats, PDF, HTML, SRT, InDesign and Contentful JSON. For a shop system like the one my store runs on, I found no connector documented, and I found no public information on Git or CI sync, SDKs, a CLI or webhooks.

Google Cloud Translation

There is a documented REST API with client libraries, published quotas, and document translation in and out of Cloud Storage for Word, PowerPoint, Excel and PDF files. For me the gap is decisive: on the captured pages I found no Shopify or other shop connector, no Git sync, no website JavaScript snippet and no software formats like XLIFF or JSON, so with no developer on staff I could not wire this to my store myself.

DeepL Pro & API

Pro texts are never used for model training and are deleted after processing, with only a narrow 72-hour debugging exception, and the engine is DeepL's own, so no third-party machine translation provider sits in the chain. TLS and advanced encryption, bring-your-own-key, audit logs, granular roles, SSO over SAML and OIDC, ISO 27001, SOC 2 Type II with its scope stated and BSI C5 are all documented, and the data processing agreement is downloadable from the Trust Centre. What keeps it from the top: I found no public information on a private or dedicated deployment option, and the free API tier reserves a right to perpetual storage of content — a reason to stay on paid plans.

Google Cloud Translation

The data usage page states plainly that text is held briefly in memory only and not stored, is not used to train Google's translation features, is not shared with any third party, and that Google claims no ownership in my content, while glossaries are encrypted with fine-grained access control under the Cloud Data Processing Addendum. I found no statement of certifications, audit logs or customer-set retention on the pages I read, and the Gemini-based service carries an 18-and-over restriction.

DeepL Pro & API

Translation memories export and import as TMX through the API — the export runs as a background job with a short-lived download link — glossaries can be built from uploaded files, and any content DeepL was asked to keep is deleted 90 days after the agreement ends. TMX is the only format on both the export and import side, and I found no public information on exporting glossaries, complete projects, comments or history, or on a self-hosted edition. My translation memory would travel with me; the terminology lists look like a manual job.

Google Cloud Translation

Translated documents come back as downloadable files, and glossaries can be imported in the open TMX, CSV or TSV formats. I found no documented export of translation memory or glossaries, no migration path from another tool, and nothing about what happens to my content on termination.

DeepL Pro & API

DeepL SE is the contracting entity, registered in Cologne with a German VAT ID, and the privacy policy states documents are translated exclusively on DeepL's own infrastructure in the EEA — reassuring for a Munich shop. Subprocessors are named with addresses, including Stripe in Dublin, Payrails in Berlin and Cloudflare in San Francisco for the website CDN, and ISO 27001, SOC 2 Type II and BSI C5 certifications plus a published data processing agreement all appear on the vendor's own pages. The one line I would want clarified before signing: the license terms reserve processing on third-party cloud infrastructure with DeepL determining the location of data processing.

Google Cloud Translation

The legal notice names Google Ireland Limited in Dublin with an Irish register number and VAT ID, and a published Cloud Data Processing Addendum covers the service. The Advanced API can pin data location with regional endpoints, but global endpoints process data with no region guarantee, the parent is Google LLC in California, and I found nothing on the captured pages about the hosting infrastructure or where the engines and subprocessors sit.

DeepL Pro & API

I can nearly compute my real annual cost from public pages alone: Individual at €7.49, Team at €24.99 and Business at €49.99 per user/month billed annually, the Write Pro add-on at €6.99 per user, and the API Growth plan at €23.80 per month plus €22.00 per extra million characters, all with VAT included and a stated 20 % saving on annual billing. Volumes per tier — characters per user, file translations, glossaries, seats, and monthly caps on the API — are listed, so a growing catalogue does not silently double my bill on Business, which shows unlimited characters. Two caveats: that "unlimited" sits under a fair usage policy whose limits are not defined, and Enterprise is contact-sales only.

Google Cloud Translation

I can genuinely compute a monthly bill from the public pages: the first 500,000 characters are free, then $20 per million characters for the NMT model, $10 per million characters in and out for the LLM model, $25 per million characters for adaptive translation, and $0.08 per page for document translation, plus $300 in free credits for new customers and no automatic charges. What stayed unclear is VAT treatment, the price beyond the top published character bands, and what the one- and three-year discounted plans commit me to.