PDF Translation Showdown: DeepL vs Google vs Microsoft in 2026

PDF Translation Showdown: DeepL vs Google vs Microsoft in 2026

Key takeaways

TakeawayDetail
DeepL Pro costs €20–€40/month for 5–20 PDFsPaid plans cap at 20 PDFs/month (20 MB each), with overages at €0.02/MB.
Google charges ~$0.02/page for PDFsPay-as-you-go pricing at $0.00002/character, with a 500K-character free tier.
Microsoft’s free tier excludes PDFsRequires copy-paste or paid API ($10/month for 1M characters).
Google supports 133 languages vs DeepL’s 32Includes rare languages like Hawaiian and Yiddish, missing in competitors.
DeepL fails on RTL/scripts (Arabic, Thai)Google and Microsoft handle complex scripts; DeepL only supports Latin-based languages.
Microsoft preserves formatting best for tables/fontsOutperforms Google but trails DeepL for European languages.
No tool retains hyperlinks/bookmarksDeepL loses all interactive elements; Google preserves hyperlinks only.
Microsoft allows 50 MB PDFs; DeepL caps at 20 MBGoogle’s web interface maxes at 10 MB; Microsoft splits large files.

Useful thresholds

ItemRule / threshold
Free-tier PDF limitDeepL: 3 PDFs/month (5 MB each); Google: 10 MB/file (no page limit); Microsoft: 0 PDFs (text only)
Paid plan entry costDeepL: €20/month; Google: $0.00002/character; Microsoft: $10/month
Max file sizeDeepL: 20 MB (paid), 5 MB (free); Google: 10 MB; Microsoft: 50 MB
"Good deal" threshold<5 PDFs/month: DeepL free tier; 5–20 PDFs: DeepL Pro; >20 PDFs: Google/Microsoft API
Overage feesDeepL: €0.02/MB; Google: $0.00002/character; Microsoft: $0.00001/character

PDF Translation Showdown: DeepL vs Google vs Microsoft in 2026

Tool Best For Cost (10-Page PDF) Max File Size Languages (PDF) Formatting Preservation
DeepL Pro European languages, simple layouts €0.20 (included in plan) 20 MB (paid) 32 Excellent (simple layouts)
Google Translate Rare languages, scanned PDFs $0.20 (≈$0.00002 × 10,000 characters for 10 pages) 10 MB (web interface), 100 MB (API only) 133 Good (hyperlinks retained)
Microsoft Translator Complex tables, HIPAA compliance $0.10 (≈$0.00001 × 10,000 characters for 10 pages) 50 MB (API, splits >10 MB) 125 Very Good (multi-column)

How Much Does Each Tool Cost for PDF Translation in 2026?

Current pricing as of July 2026 varies by volume and use case. DeepL Pro charges flat monthly fees (20 for 5 PDFs, 40 for 20 PDFs) with overages at 0.02/MB. Google Cloud Translation API uses pay-as-you-go character pricing ($0.00002/character ≈ $0.02/page). Microsoft Azure AI Translator starts at $10/month for 1M characters ($0.00001/character for overages).

For a standard 10-page PDF (≈2 MB, 1,000 characters/page):

  • DeepL Pro: €0.20 (included in €20/€40 plans)
  • Google: $0.20 (500,000-character free tier covers ≈500 pages)
  • Microsoft: $0.10 (500,000-character free tier for text-only; PDFs require paid API)

Key differences:

  • DeepL’s free tier allows 3 PDFs/month (5 MB each). Google’s free web interface has no page limit but caps at 10 MB. Microsoft’s free tier excludes PDF uploads.
  • Google’s API costs $0.00002/character vs Microsoft’s $0.00001/character but includes OCR for scanned PDFs. Microsoft’s $10/month plan covers 1M characters (≈1,000 pages at 1,000 chars/page), making it the cheapest for high-volume use.
  • DeepL’s €40/month plan is cost-effective for ≤20 PDFs/month with predictable sizes. Overage fees apply (e.g., a 12 MB PDF costs €40 + €0.04).

Common mistakes:

  • Exceeding DeepL’s 5-PDF limit on the €20 plan (triggers €0.02/MB overages).
  • Miscalculating Google’s free-tier limit (500,000 characters ≈ 500 pages, not PDFs).
  • Assuming Microsoft’s free tier supports PDFs (it’s text-only).
  • Dense pages (2,000+ characters) cost ≈$0.04/page (vs $0.02/page for 1,000 characters).
Tool Plan Cost per 10-Page PDF Max File Size Free Tier (Pages/Month)
DeepL Pro €20/month €0.20 (included) 20 MB 3 PDFs (5 MB each)
DeepL Pro €40/month €0.20 (included) 20 MB N/A
Google Cloud API Pay-as-you-go $0.20 10 MB (web interface), 100 MB (API only) 500 pages
Microsoft Azure $10/month $0.10 50 MB (API) 500 pages (text-only)

Which Tool Has the Best Free Tier for PDFs?

Google Translate offers the most generous free tier in 2026: 500,000 characters/month (≈500 pages) with a 10 MB file-size cap. DeepL’s free tier allows 3 PDFs/month (5 MB each), while Microsoft’s excludes PDF uploads entirely.

Key limitations:

  • Google: Free tier resets monthly. Mixed-language PDFs may deplete the character limit faster (each segment counts separately).
  • DeepL: 3-PDF limit does not roll over; blocked in Russia. Overage fees apply (e.g., a 6 MB PDF costs €0.02).
  • Microsoft: Free tier applies only to pasted text (500,000 characters/month). PDF uploads require the paid API.

Common pitfalls:

  • Google’s 10 MB cap fails silently for larger files.
  • DeepL’s free tier counts each PDF separately (e.g., 3 PDFs = 3 translations, regardless of size).
  • Microsoft’s web interface rejects PDFs outright.
Tool Free Tier PDF Limit File Size Cap OCR Support Reset Frequency
Google Translate 500,000 chars/month (≈500 pages) 10 MB Yes (paid API) Monthly
DeepL 3 PDFs/month 5 MB No Monthly
Microsoft Translator 0 PDFs (text-only) N/A Yes (paid API) N/A

Actionable advice:

  • Use Google’s free tier for ≤500 pages/month of native PDFs under 10 MB.
  • Use DeepL’s free tier for ≤3 PDFs/month with European languages.
  • Avoid Microsoft for PDFs—its free tier is text-only.
  • Pre-check file sizes: A 10-page PDF with 1,500 chars/page (15,000 total) leaves 485,000 chars for other documents.
  • For scanned PDFs, use paid OCR or third-party tools before translating to stay within free limits.

Language Support: Which Tool Covers the Most Languages?

As of July 2026, Google Translate supports 133 languages for PDF translation, Microsoft Translator 125, and DeepL 32. These counts apply only to direct PDF uploads; text-only translations may cover more languages but require manual copy-pasting.

Key differences:

  • Google: Prioritizes breadth, including rare languages (Hawaiian, Yiddish, Esperanto) and right-to-left scripts (Arabic, Hebrew). Free web interface supports PDFs in 64 of 133 languages.
  • Microsoft: Excludes some low-resource languages but includes niche scripts (Amharic, Hausa). Supports right-to-left scripts but may misalign text in tables.
  • DeepL: Focuses on European and major Asian languages (Japanese, Chinese). Fails on right-to-left and complex scripts (Thai, Khmer) in PDFs, though it handles them in plain text.

Exceptions:

  • Google occasionally misclassifies mixed-language segments, defaulting to English for unrecognized text.
  • Microsoft’s PDF support may drop in certain regions (e.g., China blocks Tibetan translations).
  • DeepL’s 32 languages are fully functional, but Google’s 133 include some with poor formatting retention (e.g., Burmese, Lao).
Tool Languages (PDF) Right-to-Left Scripts Rare Languages Mixed-Language Support
DeepL Pro 32 No No Weak
Google Translate 133 Yes Yes (Hawaiian, Yiddish) Strong
Microsoft Translator 125 Yes No Moderate

Decision rule:

  • Use DeepL for European languages or Japanese.
  • Use Google for rare or mixed-language PDFs.
  • Use Microsoft for right-to-left scripts or enterprise compliance.
  • Pre-check support with a 1-page test file.

Formatting Preservation: Which Tool Keeps Your PDF Intact?

No tool preserves 100% of PDF formatting, but DeepL Pro excels for European-language documents with simple layouts, Microsoft Translator for complex tables and multi-column documents, and Google Translate for mixed-language or scanned PDFs.

Key findings:

  • DeepL: Retains tables, headings, and line breaks for German, French, and Spanish but fails on right-to-left scripts (outputs garbled text). Drops hyperlinks, bookmarks, and interactive form fields. Embedded fonts are replaced with system fonts, altering page count.
  • Microsoft: Preserves column alignment and table borders better than Google but splits files >10 MB into batches, breaking page continuity. Retains form field labels but not fillable functionality. Handles low-resolution scanned PDFs well.
  • Google: Retains hyperlinks and basic paragraph structure but misaligns multi-column layouts and shifts image captions. Best for scanned PDFs (built-in OCR) and mixed-language content.

Edge cases:

  • All tools fail on password-protected PDFs—remove protection before uploading.
  • Scanned PDFs: DeepL requires third-party OCR; Google and Microsoft include OCR but may shift text overlays.
  • Embedded objects: DeepL and Microsoft reflow text, altering page count; Google substitutes fonts aggressively.
Feature DeepL Pro Google Translate Microsoft Translator
Layout preservation (simple) Excellent Good Very Good
Layout preservation (multi-column) Good Fair Very Good
Hyperlinks retained No Yes No
Built-in OCR No Yes Yes
Right-to-left scripts Fails Works Works
Max file size (paid) 20 MB 100 MB (API) 50 MB (splits >10 MB)

Decision rule:

  • Use DeepL for ≤20-page European-language PDFs with simple layouts and no hyperlinks.
  • Use Microsoft for documents with complex tables, right-to-left scripts, or file sizes up to 50 MB (accept batch-splitting delays).
  • Use Google for scanned PDFs, mixed-language documents, or when hyperlinks must survive.
  • Test a single page first—if output shifts more than one line of text, switch tools.

Maximum PDF File Size Limits in 2026

The largest PDF file you can upload for translation in 2026 is 50 MB on Microsoft Azure AI Translator, 20 MB on DeepL Pro, and 10 MB on Google Translate’s web interface. Larger files require splitting or compression.

Key details:

  • Microsoft: 50 MB cap (API-only) splits files >10 MB into batches, doubling turnaround time. Free web interface rejects PDFs.
  • DeepL: 20 MB limit (paid), 5 MB (free). Overage fees apply (€0.02/MB).
  • Google: 10 MB web limit; API supports 100 MB. Free tier includes OCR for scanned PDFs.

Exceptions:

  • Password-protected PDFs are rejected by all tools—remove protection first.
  • DeepL’s 20 MB cap is strict; overages cost €0.02/MB (e.g., 21 MB = €40 + €0.02).
  • Google’s 10 MB web limit fails silently.
Tool Max File Size (Web) Max File Size (API) Splits Files >10 MB? OCR Included?
DeepL Pro 20 MB (paid), 5 MB (free) N/A No No
Google Translate 10 MB 100 MB No Yes (paid API)
Microsoft Translator 0 (web rejects PDFs) 50 MB Yes Yes

Action rule:

  • Use Google’s free web interface for ≤10 MB.
  • For 10–20 MB, use DeepL Pro (paid) or Google’s API.
  • For 20–50 MB, use Microsoft’s API.
  • Remove password protection and test a 1-page sample.
  • For files >50 MB, split or compress with tools like Adobe Acrobat.

Regional Speed Differences: Which Tool Is Fastest in 2026?

Processing speeds vary by region in 2026. DeepL averages 30 seconds for a 10-page PDF in Frankfurt, 45 seconds in Tokyo, and 60 seconds in São Paulo. Google Translate ranges from 45–60 seconds globally. Microsoft Translator processes fastest in the U.S. East (20 seconds) and slowest in India (40 seconds).

Key factors:

  • DeepL: Dedicated clusters in Frankfurt, Tokyo, New York, São Paulo, Sydney, and Mumbai ensure consistency in core markets. Latency increases in regions with fewer clusters (e.g., Africa).
  • Google: Broad edge network (Google Cloud CDN) reduces variability but may slow during ISP throttling. Free web interface adds 5–10 seconds for ad loading.
  • Microsoft: Enterprise peering accelerates North America/Europe. Files >10 MB split into batches, doubling turnaround time. Free tier excludes PDFs.

Exceptions:

  • China: All tools route through local partners, adding 2–3 seconds.
  • Peak hours (8–10 AM local time): Speeds may drop by 30–50%.
  • Scanned PDFs: Google and Microsoft add 10–15 seconds for OCR; DeepL fails without pre-processing.

Decision rule:

  • Use DeepL for European or Japanese PDFs (fastest in Frankfurt/Tokyo).
  • Use Microsoft for North American documents (U.S. East cluster).
  • Use Google for global consistency or scanned PDFs.
  • Avoid translating during peak hours (8–10 AM local time).
  • For files >10 MB, expect Microsoft’s batch-splitting delays (adds 20–30 seconds).

Data Privacy: Which Tool Is Safest for Sensitive PDFs?

DeepL is the safest for sensitive PDFs in 2026, as it does not store content after translation. Google and Microsoft retain data for 30 days unless users opt out (Google) or use a private endpoint (Microsoft).

Key differences: