DeepL Pro & API vs Google Cloud Translation

DeepL Pro & API

Their own neural engine covers a published list of 126 languages with glossaries applied during translation, a formal-or-informal tone setting, and document translation that keeps the original layout for Word, PowerPoint, PDF and even InDesign and image files. Adaptive translation confined to our own account, style rules with custom instructions, and a glossary generator that works from at least 1,000 sentence pairs of our past translations push toward the next level. We found no public information on per-segment quality estimation or a published quality evaluation with its method, and the engine is DeepL's own rather than a choice of engines per language pair.

Google Cloud Translation

Three engines to choose from — Google's neural model, a translation large language model with contextual glossary support, and adaptive translation that learns from our own small datasets without training — covering 189 published languages, plus AutoML custom models for 53 language pairs to and from English, and document translation that preserves the layout for Word, PowerPoint and PDF. Glossaries are applied during translation with case handling, term-set types and TMX import, which is what my termbase needs. We found no public information on a published quality evaluation with its method, and quality prediction appears only as a Translation Hub tier feature without routing of weak segments into review.

DeepL Pro & API

A team-shareable translation memory (TMX upload, reuse of similar past translations) sits beside centrally managed, mergeable glossaries with priority ordering, style rules and shareable style profiles — the terminology backbone I need is genuinely there. But we found no public information on review or approval stages, automated quality checks that stop a broken placeholder before it ships, per-language roles, or inviting translators and agencies with cost and progress per vendor; the pages instead show CAT tool integration for the paid API, so any review workflow would live outside DeepL.

Google Cloud Translation

Glossaries are managed centrally with create, list, get, update and delete operations, and the separately priced Translation Hub advertises translation memory, human review and machine translation quality prediction scores in its advanced tier — but glossary version control is explicitly not supported. We found no public information on automated checks for placeholders and tags, per-language reviewer roles, an in-context editor, or inviting agencies and freelancers into the product, so I could not run a fourteen-language release review on it. A terminology store plus one review stage for documents, not a managed process.

DeepL Pro & API

The REST API is documented in detail, down to translation-memory endpoints with signed upload URLs and explicit status codes, and connectors cover Zendesk, Adobe Experience Manager, SharePoint, Google Drive, OneDrive, Microsoft 365, Google Workspace, AWS Marketplace and browser extensions for Chrome, Edge and Firefox. Document and office formats are well covered (PDF, Word, PowerPoint, Excel, HTML, SRT, TXT, InDesign), with Contentful JSON the only software format named. We found no public information on XLIFF, PO, YAML, Android or iOS string formats, Git repository sync, a CLI, SDKs, webhooks or a public status page — a gap for me since my product strings are JSON and strings files, not slide decks.

Google Cloud Translation

A documented REST API with client libraries, batch jobs reading from and writing to Cloud Storage, published quotas and limits, and document translation for Word, PowerPoint, Excel and PDF give my engineers everything for custom pipelines. We found no public information on the software localization formats I ship in — XLIFF, PO, JSON, iOS and Android strings — nor on Git or CI synchronisation, CMS, shop or design connectors, or over-the-air delivery. A strong API to build on, but nothing that plugs into our repositories and CMS.

DeepL Pro & API

For Pro, texts are stored only as technically needed and deleted after the service, never used to train their models — with a 72-hour debugging exception and deletion of customer-requested storage 90 days after the agreement ends — plus bring-your-own-key encryption, audit logs, role-based permissions, SSO over SAML and OpenID Connect, ISO 27001 and SOC 2 Type II with its scope stated, and a data processing agreement forming part of the terms. The free tiers read differently: free-service texts feed model training and the free API reserves perpetual storage of content, so the paid plans are where these guarantees live. We found no public information on a private or dedicated deployment option.

Google Cloud Translation

The data usage statements are unusually firm: sent text is held briefly in-memory only and not stored on servers, is not used to train Google's translation features, is never shared with third parties or made public, Google claims no ownership, and the service complies with the Cloud Data Processing Addendum, with glossaries encrypted and guarded by fine-grained IAM permissions. We found no public information on ISO 27001 or SOC 2 certification with its scope, audit logs, or single sign-on, and encryption in transit and at rest is not stated for translated content. Contractually clean, but nothing on the pages lets me verify it independently.

DeepL Pro & API

Translation memories export as TMX through a documented background-job API and import the same way — TMX is stated as the only supported format in either direction — and account deletion is self-service, with customer-requested stored content deleted 90 days after the agreement ends. We found no public information on glossary export in TBX or CSV, a bulk export of all projects and files, or any self-hosted or on-premise edition, so the exit path today runs through the translation memory alone.

Google Cloud Translation

My terminology can leave: glossaries import as TSV, CSV or TMX from my own Cloud Storage bucket, every entry can be listed back through the API, and translated documents are returned to me or written to a bucket. We found no public information on exporting a translation memory as TMX, a complete project export, a migration path from another TMS, data return or deletion on termination, or a self-hosted edition. Assets are recoverable one by one, not as a documented exit package.

DeepL Pro & API

DeepL SE is registered in Cologne (Amtsgericht Köln, HRB 104617), carries ISO 27001, SOC 2 Type II and BSI C5 Type 2 certifications, publishes a data processing agreement as part of its terms, and names subprocessors with addresses: Stripe in Dublin, Payrails in Berlin, Usercentrics in Munich, Cloudflare in San Francisco, Apryse for the in-house PDF translation SDK and ElevenLabs for backup voices. The captured pages give differing indications on where content runs — the privacy policy states document translation is processed exclusively on DeepL's infrastructure in the EEA, while the terms reserve processing on third-party cloud infrastructure with DeepL determining the location. With a US CDN and PDF SDK in the chain and no location given for two subprocessors, this is a strong European seat that stops short of a fully evidenced end-to-end chain.

Google Cloud Translation

Google Ireland Limited in Dublin is the named service provider with an Irish registration and VAT number, the Cloud Data Processing Addendum is published with GDPR commitments described, and the Advanced API offers regionalized endpoints configurable for data location — though the Basic endpoints are explicitly global and cannot be region-locked. The parent company is in the USA, and we found no public information on hosting regions, subprocessors with their locations, or security certifications. Enough EU entity and contractual footing for procurement, not enough to show my data protection officer the full chain.

DeepL Pro & API

Seat prices are public per tier — Individual €7.49, Team €24.99, Business €49.99 per user/month billed annually — with included characters, file translations and glossary counts listed, and the API side is unit-priced: €23.80 per month billed annually plus €22.00 per 1,000,000 extra characters, €1.96 per extra speech-to-text hour and €3.93 per extra speech-to-speech hour, with 20 % annual discount and VAT stated as included. Enterprise is custom pricing only and the fair usage policy behind "unlimited characters" is not quantified, and the captured pages give different indications on SSO — plan pages list it for teams of 50+ users while the SSO page says contact sales — so the complete annual cost for a large organisation still needs a sales call.

Google Cloud Translation

From the public pages alone I can compute the annual bill: $20 per million characters for the neural model with the first 500,000 characters per month free, $10 per million for the LLM, $25 per million for adaptive, custom translation at $80 per million with volume bands down to $30, document translation at $0.08 or $0.25 per page, Translation Hub at $0.15 or $0.50 per page per target language, custom model training at $45 per hour capped at $300 per job, optional one- and three-year commitment prices, and $300 in free credits with no commitment. We found no public information on VAT treatment, and no self-hosted edition to price. This is the most computable price list I have graded.