Detect Language of Text

Very accurate for sentences and paragraphs (95%+ accuracy). The tool uses statistical n-gram analysis (franc library) which examines character patterns and frequencies. Accuracy depends on text length: Paragraphs (100+ words): 95-99% accurate. Sentences (10-50 words): 85-95% accurate. Short phrases (3-9 words): 60-80% accurate (ambiguous). Single words: Often inaccurate (insufficient context). The confidence score indicates certainty—90%+ means highly confident, below 70% suggests ambiguity. Always check the top 3 candidates when confidence is low.

Confidence score (0-100%) indicates statistical certainty that the detected language is correct. 90-100%: Extremely confident—text strongly matches this language's patterns. 70-89%: Confident—likely correct but some ambiguity exists. 50-69%: Moderate confidence—could be correct or a close relative language. Below 50%: Low confidence—text is too short, mixed languages, or an unusual dialect. When confidence is below 80%, check the Top 3 Candidates—the actual language might be ranked #2 or #3, especially for closely related languages (e.g., Spanish vs. Portuguese, Hindi vs. Urdu).

The Top 3 Language Candidates show the most likely languages ranked by confidence, with visual progress bars indicating relative likelihood. Useful when: (1) Primary detection has low confidence (<80%)—check if another candidate seems more plausible. (2) Text mixes languages—see which languages are present. (3) Closely related languages (Spanish/Portuguese, Hindi/Urdu)—compare candidates to determine the correct one. (4) Learning: Understand which languages share similar character patterns. Example: Input 'Olá' (Portuguese) might show: Portuguese 75%, Spanish 68%, Galician 45%. Compare candidates with your knowledge of the source.

82+ languages covering the vast majority of written internet content. Major Languages: English, Spanish, French, German, Italian, Portuguese, Russian, Japanese, Mandarin Chinese, Arabic, Hindi, Korean. European: Polish, Dutch, Swedish, Czech, Romanian, Greek, Ukrainian, Norwegian, Danish, Finnish, Hungarian, Croatian, Serbian, Bulgarian. Asian: Bengali, Tamil, Telugu, Gujarati, Marathi, Kannada, Thai, Vietnamese, Indonesian, Malay, Tagalog, Khmer, Burmese, Mongolian. Middle Eastern: Hebrew, Persian, Urdu, Pashto, Kurdish, Turkish, Azerbaijani. African: Swahili, Amharic, Afrikaans. Plus many more dialects and regional languages.

Statistical n-gram analysis requires sufficient character patterns to identify language reliably. Why short text is problematic: (1) Insufficient data points—single words don't have enough trigrams (3-character sequences), (2) Ambiguity—'Hola' could be Spanish, but also appears in other Romance languages, (3) Loanwords—'Café' appears in English, French, Portuguese, Spanish, (4) Proper nouns—names and places don't indicate language. Solutions: Add more context (full sentences), provide surrounding text, check top 3 candidates when confidence is low, use knowledge of the source to disambiguate. Rule of thumb: Minimum 15-20 characters (2-3 words) for reasonable accuracy.

Content Moderation: Auto-route foreign language comments to appropriate moderation queues or translation services. Translation Prep: Identify source language before sending to Google Translate, DeepL, or human translators. Data Processing: Classify multilingual datasets, user feedback, or survey responses by language for analysis. API Validation: Verify user input matches expected locale in web forms or applications (e.g., ensure 'English' content in English-only fields). Language Learning: Test your ability to identify languages or discover what language you're reading. SEO/Content Strategy: Detect language of scraped content for localization or competitive analysis.

Partially. The tool identifies the dominant language in mixed text but doesn't segment multiple languages. Example: 'Hello everyone, cómo están?' → Likely detects English (dominant). However, check Top Candidates to see if Spanish appears with moderate confidence, indicating mixed content. Workaround for mixed text: Split text manually by language, detect each section separately. Code-switching (alternating languages mid-sentence) will detect whichever language has more characters/words. For true multilingual segmentation (splitting text into language blocks), use specialized polyglot detection libraries not available in browser-based tools.

ISO 639-3 is an international standard for 3-letter language codes uniquely identifying each language. Unlike 2-letter codes (ISO 639-1), 3-letter codes cover 7,000+ languages including rare and ancient ones. Examples: eng = English, spa = Spanish, fra = French, deu = German (Deutsch), cmn = Mandarin Chinese, ara = Arabic, jpn = Japanese, hin = Hindi. Useful for: Technical documentation, database language fields, API payloads, translation services, linguistic research. The tool displays both human-readable name AND ISO code for technical precision.

No. 100% client-side processing. The tool uses the franc library which runs entirely in your browser using JavaScript. Your text never leaves your device, isn't uploaded to our servers or any third party, isn't logged, and isn't stored anywhere. Check your browser's Network tab to verify zero external requests during detection. This is critical for: Confidential documents, personal communications, unpublished manuscripts, proprietary content, sensitive business data, or any text requiring complete privacy. Unlike cloud-based APIs (Google, Microsoft, AWS), this tool provides instant results without any data transmission.

Yes, but with manual copy/paste currently. The tool offers: File Upload: Process .txt/.md files directly. Download Report: Save detection results as .txt with full statistics and top candidates. Copy Result: Quick copy of detected language for pasting elsewhere. For true automation (API-like usage), you would need to: (1) Use the franc library directly in your Node.js/JavaScript project, (2) Call cloud APIs (Google Cloud Translation API, Azure Text Analytics), (3) Use command-line tools like cld2/cld3. This browser tool is optimized for manual ad-hoc detection with privacy as the priority, not high-volume batch processing.