DeepL

DeepL is the translation engine that professional translators use alongside clients — its neural machine translation consistently outperforms Google Translate and Microsoft Translator on European language pairs in independent evaluations. For indie authors, the practical applications are metadata, blurbs, back-cover copy, and marketing materials where quality matters but full manuscript literary translation isn't the priority.

What You Get#

• 100+ languages (1,000+ translation combinations): Among the broadest coverage of any AI translation tool — roughly 30 core languages run on both the classic and next-generation models, while ~100 additional languages (Hindi, Thai, Vietnamese, Hebrew, Welsh, and others) are supported via the next-gen model only
• Document translation: Upload PDF, Word (.docx), or PowerPoint files and receive a translated version with formatting preserved — no reformatting required
• Glossary support: Define how specific terms, character names, or branded phrases should be translated consistently across a project — essential for fiction series with unique terminology
• DeepL Write: A companion writing enhancement tool for improving and paraphrasing text in the target language (available on paid plans)
• API access: Available separately for developers or authors integrating translation into automated workflows and publishing pipelines

Who It's For#

• Authors targeting European markets: DeepL's neural models are strongest on European language pairs — Spanish, French, German, Italian, Portuguese, Dutch, Polish, and Russian see the best accuracy
• Authors translating metadata and marketing copy: Cost-effective for translating book descriptions, blurbs, author bios, keywords, and ad copy without paying for full literary translation
• Authors using professional CAT tools: Integrates with translation memory and terminology tools for hybrid human+AI workflows where a human translator finalizes the output

Pricing: DeepL restructured its consumer plans in 2026, retiring the old Starter/Advanced/Ultimate tiers. The current lineup is Free, Individual (~$8.74/month), Team (~$28.74/user/month), Business (~$57.49/user/month), and Enterprise (custom). The Free plan covers 50,000 characters/month (~8,300 words) and 1 document/file translation — enough for back-cover copy, an author bio, and keyword metadata, but tight for heavier workflows. Paid tiers add far higher character limits, more document translations, larger glossaries, CAT-tool integration, and a no-training data guarantee. The quoted figures are annual-billing monthly-equivalents; monthly billing runs roughly 16% higher, and exact prices vary by region — verify on deepl.com/pro before committing. Existing customers on the legacy Starter/Advanced/Ultimate plans may be grandfathered.

Strengths#

• European language accuracy: Consistently rated highest for accuracy in European language pairs in independent head-to-head evaluations against Google Translate and Microsoft Translator
• Glossary for consistency: Ability to lock down character names, place names, and series terminology is essential for fiction translation and rare among free tools
• Document format preservation: Returns translated documents in the original file format — a practical advantage over copy-paste translation workflows
• Free tier is genuinely useful: 50,000 characters/month (~8,300 words) is enough to translate several back-cover descriptions, an author bio, and keyword metadata every month at no cost — though manuscript-length work needs a paid plan

Limitations#

• Not EPUB-native: Authors wanting to translate full manuscripts must convert to Word (.docx) first, then handle reformatting after — adding workflow steps that book-specific tools avoid
• Asian language quality gap: Performance on Chinese, Japanese, and Korean is improving but still noticeably behind the accuracy on European language pairs
• General-purpose, not literary: The neural model optimizes for fluency and accuracy, not voice preservation — a fiction author's stylistic choices, rhythm, and register will be flattened without human post-editing
• Not suitable for production-ready literary translation: Treating DeepL output as a finished product rather than a first draft produces translations that will feel mechanical to native readers
• Free-tier data is used for model training: Text run through the Free plan may be used to improve DeepL's models — a privacy concern for unpublished manuscripts or confidential material. The no-training guarantee applies only to paid plans

Alternatives#

For full manuscript translation purpose-built for books, BookTranslate.ai and BookTranslator.ai are AI tools that handle EPUB natively. For professionally reviewed human literary translation, Espresso Translations and Reedsy's translator marketplace provide native-speaking specialists with quality review processes.