MAITO Logo MAITO Logo
Knowledge Base
Back to Knowledge Base

Language Support and Translation Capabilities

Complete guide to languages supported by MAITO's DeepL and Local AI engines, including language-specific features and limitations

Pangaia Software • • 7 min read

MAITO supports a wide range of languages through its dual-engine system. This guide covers all supported languages and their capabilities.

Overview

  • DeepL: 30+ languages
  • Local AI (Rosetta): 32 languages
  • Auto-Detection: Supported by both engines
  • Language Pairs: Any source → any target (within engine limits)

DeepL Supported Languages

Fully Supported Languages (30+)

DeepL supports these languages for text, document, and website translation:

European Languages:

  • Bulgarian
  • Czech
  • Danish
  • Dutch
  • English (American & British)
  • Estonian
  • Finnish
  • French
  • German
  • Greek
  • Hungarian
  • Italian
  • Latvian
  • Lithuanian
  • Polish
  • Portuguese (European & Brazilian)
  • Romanian
  • Slovak
  • Slovenian
  • Spanish
  • Swedish

Asian Languages:

  • Chinese (Simplified)
  • Japanese
  • Korean

Other:

  • Indonesian
  • Norwegian (Bokmål)
  • Russian
  • Turkish
  • Ukrainian

DeepL Updates

DeepL regularly adds new languages. Check the DeepL website for the most current list. MAITO automatically supports new DeepL languages as they’re released.

DeepL Language-Specific Features

Formality Control: Available for: German, French, Italian, Spanish, Dutch, Polish, Portuguese, Russian, Japanese

Variants:

  • English: American vs British
  • Portuguese: European vs Brazilian
  • Chinese: Simplified only (Traditional not supported)

Local AI (Rosetta) Supported Languages

All Supported Languages (32)

Rosetta models support these languages:

European Languages:

  • Bulgarian
  • Czech
  • Dutch
  • English
  • French
  • German
  • Greek
  • Hungarian
  • Italian
  • Polish
  • Portuguese
  • Romanian
  • Russian
  • Slovak
  • Spanish
  • Swedish
  • Ukrainian

Asian Languages:

  • Chinese (Simplified)
  • Chinese (Traditional)
  • Gujarati
  • Hindi
  • Japanese
  • Korean
  • Persian
  • Tagalog
  • Vietnamese

Middle Eastern:

  • Arabic
  • Hebrew
  • Turkish

Other:

  • Indonesian

Local AI Limitations

No Formality Control: Rosetta produces neutral tone only

No Website Translation: webpages still require DeepL

Partial Document Support: Word (.docx, .doc), PDF and TXT files are translated offline; PowerPoint and HTML require DeepL. Right-to-left languages such as Arabic and Hebrew are handled correctly in Word documents.

Language Comparison

Languages in Both Engines

These languages work in both DeepL and Local AI:

  • Bulgarian, Chinese (Simplified), Czech, Dutch, English
  • French, German, Greek, Hungarian, Indonesian
  • Italian, Japanese, Korean, Polish, Portuguese
  • Romanian, Russian, Slovak, Spanish, Swedish
  • Turkish, Ukrainian

Total Overlap: 22 languages

DeepL Exclusive

  • Danish, Estonian, Finnish, Latvian
  • Lithuanian, Norwegian, Slovenian

Local AI Exclusive

  • Arabic, Chinese (Traditional), Gujarati, Hebrew
  • Hindi, Persian, Tagalog, Vietnamese

Best of Both Worlds

Configure both engines to access the widest language coverage - 32 unique languages total across both systems!

Auto-Detection

How It Works

Both engines can automatically detect the source language:

  1. You select “Auto Detect” as source language
  2. Enter text to translate
  3. Engine analyzes text and identifies language
  4. Translation proceeds to target language

Accuracy

DeepL Auto-Detection:

  • Very accurate for supported languages
  • Requires minimum 10-20 characters
  • Mixed language text may confuse detection

Local AI Auto-Detection:

  • Good accuracy for supported languages
  • Requires minimum 20-50 characters
  • Better with complete sentences

When Auto-Detection Fails

Short Text:

  • “Hello” → May detect as multiple languages
  • Add more context for better detection

Mixed Languages:

  • “Hello, comment ça va?” → Confusion between English/French
  • Manually select source language

Similar Languages:

  • Czech vs Slovak
  • Portuguese European vs Brazilian
  • Manually select for consistency

Best Practices

Use Auto-Detect When:

  • Source language unknown
  • Receiving text from varied sources
  • Convenient for workflow

Manually Select When:

  • Translating specific language regularly
  • Need consistent language pair
  • Auto-detect gives wrong results

Translation Quality by Language Pair

DeepL Quality

Excellent Quality (Major European Languages):

  • English ↔ German, French, Spanish, Italian
  • These pairs have the most training data
  • Near-human quality in many cases

Very Good Quality (Other Supported Languages):

  • All other DeepL language pairs
  • May have minor phrasing differences
  • Excellent for comprehension

Local AI (Rosetta) Quality

Optimized Pairs (Most Common):

  • English ↔ German, French, Spanish, Chinese, Japanese
  • Most training data and optimization

Well-Supported (All 32 Languages):

  • All Rosetta language pairs work well
  • Some pairs may be slightly less polished
  • Excellent for general understanding

Best For:

  • English ↔ Any supported language
  • Romance languages (FR, ES, IT, PT) ↔ Each other
  • Major Asian languages (ZH, JA, KO) ↔ English

Special Language Considerations

Chinese

Simplified vs Traditional:

  • DeepL: Simplified only
  • Local AI: Both Simplified and Traditional
  • Traditional Chinese users: Use Local AI

Variants: Both mainland and Taiwan usage supported in Local AI

Portuguese

European vs Brazilian:

  • DeepL: Separate selections for PT-PT and PT-BR
  • Local AI: Unified Portuguese (works for both)

Recommendation: DeepL if you need specific variant, Local AI for general Portuguese

English

American vs British:

  • DeepL: Can specify target variant
  • Local AI: Produces neutral English (mostly American)

Spelling:

  • DeepL: Honors variant (color vs colour)
  • Local AI: Typically American spelling

Arabic

Only in Local AI: Not supported by DeepL

Right-to-Left: Properly handled by MAITO UI

Dialects: Modern Standard Arabic (MSA) primarily

Japanese

Kanji, Hiragana, Katakana: All properly handled

Formality:

  • DeepL: Formality control available (polite vs casual)
  • Local AI: Neutral tone

Romanization: Not supported (use Japanese characters)

Korean

Hangul: Properly supported

Formality Levels: Neutral tone (Local AI), contextual (DeepL)

Romanization: Not supported (use Hangul)

Language Pair Performance

Translation Speed by Language

DeepL: Consistent speed for all language pairs (cloud-based)

Local AI: Varies slightly by language

Fastest (Local AI):

  • English ↔ European languages
  • Chinese ↔ English

Slightly Slower (Local AI):

  • Complex scripts (Arabic, Hebrew)
  • Languages with large character sets (Chinese, Japanese)

Difference: Typically 10-20% variation, rarely noticeable

Language-Specific Limitations

DeepL

Character Limits by Language:

  • All languages: Up to 130,000 characters per request
  • Consistent across all supported languages

Document Support:

  • All supported languages work with documents
  • Formatting preserved equally well

Local AI (Rosetta)

Text Length:

  • No practical limit for text translation
  • Intelligent chunking at sentence/paragraph boundaries
  • Completely transparent to users

Context Window:

  • Full: 16,384 tokens (~65,000 chars)
  • May be reduced on systems with <6GB RAM
  • Affects very long text translation

Multilingual Workflows

Translate to Multiple Languages

Scenario: Translate English to multiple targets

DeepL Approach:

  1. Translate to Language 1
  2. Change target language
  3. Translate again to Language 2
  4. Repeat as needed

Local AI Approach: Same as DeepL

Tip: Keep source text in left panel, only change target language

Chain Translations

Scenario: Translate A → B → C (pivot through intermediate language)

Example: Arabic → English → Japanese

  • Arabic text → Translate to English (Local AI)
  • English result → Translate to Japanese (either engine)

When Useful: When direct pair unavailable or lower quality

Quality Loss

Chain translations lose some quality at each step. Use direct translation pairs when possible.

Language Learning Use Case

MAITO works great for language learners:

Comprehension Practice:

  • Read foreign text
  • Translate phrases you don’t understand
  • Local AI for privacy and availability (study while traveling!)

Writing Practice:

  • Write in target language
  • Translate back to native language (check meaning)
  • Compare with intended message

Vocabulary Building:

  • Translate individual words and phrases
  • See context usage
  • Build personal vocabulary lists

Regional Variants

Currently Supported

English:

  • American English
  • British English (DeepL)

Portuguese:

  • European Portuguese
  • Brazilian Portuguese (DeepL)

Chinese:

  • Simplified Chinese
  • Traditional Chinese (Local AI only)

Not Distinguished

Spanish: Castilian, Latin American (unified in both engines)

French: European, Canadian (unified in both engines)

Arabic: MSA, dialects (unified in Local AI)

Summary

Total Unique Languages: 32 across both engines Overlap: 22 languages in both engines DeepL Exclusive: 8 languages Local AI Exclusive: 10 languages

Best Coverage Strategy: Configure both engines to access maximum language support!

For specific language questions or feature requests, contact Pangaia Software at contact@pangaia.software.

📚 Related Articles

Recently Viewed