OCR Language Support

Each OCR backend supports different sets of languages. Select the backend and language based on your document requirements.

Backend Language Coverage

Tesseract

Default OCR backend with 100+ supported languages. Supports both ISO 639-3 (three-letter codes) and variant codes for historical and script-specific variants.

Language	Code	Status
Afrikaans	`afr`	Full
Amharic	`amh`	Full
Arabic	`ara`	Full
Assamese	`asm`	Full
Azerbaijani (Latin)	`aze`	Full
Azerbaijani (Cyrillic)	`aze_cyrl`	Full
Belarusian	`bel`	Full
Bengali	`ben`	Full
Tibetan	`bod`	Full
Bosnian	`bos`	Full
Breton	`bre`	Full
Bulgarian	`bul`	Full
Catalan	`cat`	Full
Cebuano	`ceb`	Full
Czech	`ces`	Full
Chinese (Simplified)	`chi_sim`	Full
Chinese (Traditional)	`chi_tra`	Full
Cherokee	`chr`	Full
Corsican	`cos`	Full
Welsh	`cym`	Full
Danish	`dan`	Full
German	`deu`	Full
Dhivehi	`div`	Full
Dzongkha	`dzo`	Full
Greek	`ell`	Full
English	`eng`	Full
Middle English	`enm`	Full
Esperanto	`epo`	Full
Equations	`equ`	Math formulas
Estonian	`est`	Full
Basque	`eus`	Full
Faroese	`fao`	Full
Persian	`fas`	Full
Filipino	`fil`	Full
Finnish	`fin`	Full
French	`fra`	Full
Frankish (Old German)	`frk`	Full
Middle French	`frm`	Full
Frisian	`fry`	Full
Scottish Gaelic	`gla`	Full
Irish	`gle`	Full
Galician	`glg`	Full
Ancient Greek	`grc`	Full
Gujarati	`guj`	Full
Haitian Creole	`hat`	Full
Hebrew	`heb`	Full
Hindi	`hin`	Full
Croatian	`hrv`	Full
Hungarian	`hun`	Full
Armenian	`hye`	Full
Inuktitut	`iku`	Full
Indonesian	`ind`	Full
Icelandic	`isl`	Full
Italian	`ita`	Full
Italian (Old)	`ita_old`	Full
Javanese	`jav`	Full
Japanese	`jpn`	Full
Kannada	`kan`	Full
Georgian	`kat`	Full
Georgian (Old)	`kat_old`	Full
Kazakh	`kaz`	Full
Khmer	`khm`	Full
Kyrgyz	`kir`	Full
Kurmanji (Kurdish)	`kmr`	Full
Korean	`kor`	Full
Lao	`lao`	Full
Latin	`lat`	Full
Latvian	`lav`	Full
Lithuanian	`lit`	Full
Luxembourgish	`ltz`	Full
Malayalam	`mal`	Full
Marathi	`mar`	Full
Macedonian	`mkd`	Full
Maltese	`mlt`	Full
Mongolian	`mon`	Full
Māori	`mri`	Full
Malay	`msa`	Full
Burmese	`mya`	Full
Nepali	`nep`	Full
Dutch	`nld`	Full
Norwegian	`nor`	Full
Occitan	`oci`	Full
Odia	`ori`	Full
Orientation/Script detection	`osd`	Layout
Punjabi	`pan`	Full
Polish	`pol`	Full
Portuguese	`por`	Full
Pushto	`pus`	Full
Quechua	`que`	Full
Romanian	`ron`	Full
Russian	`rus`	Full
Sanskrit	`san`	Full
Sinhala	`sin`	Full
Slovak	`slk`	Full
Slovenian	`slv`	Full
Sindhi	`snd`	Full
Spanish	`spa`	Full
Spanish (Old)	`spa_old`	Full
Albanian	`sqi`	Full
Serbian	`srp`	Full
Serbian (Latin)	`srp_latn`	Full
Sundanese	`sun`	Full
Swahili	`swa`	Full
Swedish	`swe`	Full
Syriac	`syr`	Full
Tamil	`tam`	Full
Tatar	`tat`	Full
Telugu	`tel`	Full
Tajik	`tgk`	Full
Thai	`tha`	Full
Tigrinya	`tir`	Full
Tonga	`ton`	Full
Turkish	`tur`	Full
Uyghur	`uig`	Full
Ukrainian	`ukr`	Full
Urdu	`urd`	Full
Uzbek	`uzb`	Full
Uzbek (Cyrillic)	`uzb_cyrl`	Full
Vietnamese	`vie`	Full
Yiddish	`yid`	Full
Yoruba	`yor`	Full

Installation: Tesseract language packs are installed via your OS package manager. Install the base Tesseract, then add individual language packs as needed. See the OCR guide for installation steps by platform.

Selection: Use --ocr-language flag with a single code (e.g., --ocr-language eng). For multiple languages, join with +: --ocr-language eng+deu+fra.

PaddleOCR

Fast text detection and recognition for 80+ languages across 11 script families. Optimized for both mobile and server deployment.

Language	Code	Family
Afrikaans	`afr`	Latin
Arabic	`ara`	Arabic
Bulgarian	`bul`	Cyrillic
Chinese (Simplified)	`ch_sim` / `zh_hans`	CJK
Chinese (Traditional)	`ch_tra` / `zh_hant`	CJK
Czech	`cs`	Latin
Danish	`da`	Latin
Dutch	`nl`	Latin
English	`en`	Latin
Estonian	`et`	Latin
Finnish	`fi`	Latin
French	`fr`	Latin
German	`de`	Latin
Greek	`el`	Greek
Hungarian	`hu`	Latin
Indonesian	`id`	Latin
Italian	`it`	Latin
Japanese	`ja`	CJK
Korean	`ko`	CJK
Latin	`la`	Latin
Latvian	`lv`	Latin
Lithuanian	`lt`	Latin
Norwegian	`nb`	Latin
Persian	`fa`	Arabic
Polish	`pl`	Latin
Portuguese	`pt`	Latin
Romanian	`ro`	Latin
Russian	`ru`	Cyrillic
Slovak	`sk`	Latin
Slovenian	`sl`	Latin
Spanish	`es`	Latin
Swedish	`sv`	Latin
Tagalog	`tl`	Latin
Turkish	`tr`	Latin
Ukrainian	`uk`	Cyrillic
Vietnamese	`vi`	Latin

Installation: Built into Xberg via the paddle-ocr feature. Models download automatically on first use.

Selection: Use --ocr-language or config ocr.languages with a single code or list. Join multiple codes with + for CLI: --ocr-language en+de+zh_hans.

Note: PaddleOCR uses two-letter ISO 639-1 codes and script-specific variants (e.g., zh_hans for Simplified Chinese). Consult the paddleocr-vl backend source for the authoritative list.

Sceptre

Sceptre implements the complete EasyOCR Gen2 recognizer family. Xberg accepts the raw model-group aliases english, latin, chinese_simplified, japanese, korean, cyrillic, telugu, and kannada. It also accepts the exact EasyOCR tokens and the additional ISO aliases below. A language is supported only when it appears here; sharing a script with a model does not make an arbitrary language code valid.

Group	Official EasyOCR Gen2 tokens	Additional aliases accepted by Xberg
`english`	`en`	`eng`
`latin`	`af`, `az`, `bs`, `cs`, `cy`, `da`, `de`, `es`, `et`, `fr`, `ga`, `hr`, `hu`, `id`, `is`, `it`, `ku`, `la`, `lt`, `lv`, `mi`, `ms`, `mt`, `nl`, `no`, `oc`, `pi`, `pl`, `pt`, `ro`, `rs_latin`, `sk`, `sl`, `sq`, `sv`, `sw`, `tl`, `tr`, `uz`, `vi`	`afr`, `aze`, `bos`, `ces`/`cze`, `cym`/`wel`, `dan`, `deu`/`ger`, `spa`, `est`, `fra`/`fre`, `gle`, `hrv`, `hun`, `ind`, `isl`/`ice`, `ita`, `kur`, `lat`, `lit`, `lav`, `mri`/`mao`, `msa`/`may`, `mlt`, `nld`/`dut`, `nor`, `oci`, `pli`, `pol`, `por`, `ron`/`rum`, `sr-Latn`/`sr_Latn`, `srp-Latn`/`srp_Latn`, `slk`/`slo`, `slv`, `sqi`/`alb`, `swe`, `swa`, `fil`, `tur`, `uzb`, `vie`
`chinese_simplified`	`ch_sim`	`zh`, `zh-CN`, `zh-Hans`, `zho`, `chi`, `chs`
`japanese`	`ja`	`jpn`, `jpn_vert`
`korean`	`ko`	`kor`
`cyrillic`	`ru`, `rs_cyrillic`, `be`, `bg`, `uk`, `mn`, `abq`, `ady`, `kbd`, `ava`, `dar`, `inh`, `che`, `lbe`, `lez`, `tab`, `tjk`	`rus`, `sr-Cyrl`/`sr_Cyrl`, `srp-Cyrl`/`srp_Cyrl`, `bel`, `bul`, `ukr`, `mon`, `tg`/`tgk`
`telugu`	`te`	`tel`
`kannada`	`kn`	`kan`

Every non-English Gen2 recognizer includes English characters. You may therefore configure English with one other group. Do not combine two distinct non-English groups in one reader.

Desktop and server builds use ONNX Runtime and download checksum-pinned models into the Hugging Face cache. Android and iOS use tract and exclude the downloader: the application must resolve packaged assets to filesystem paths and set both backend_options.model.detector_path and backend_options.model.recognizer_path. The opt-in Sceptre worker build/API requires the JavaScript host to fetch and supply the CRAFT and selected recognizer model bytes, which the Rust engine verifies against the registry digests; Sceptre is excluded from the default wasm-target bundle.

Candle TrOCR

Lightweight line-level text recognition using Microsoft’s TrOCR model. Trained primarily for English.

Language	Code
English	`eng` / `en`

Variants: Four model sizes are available:

base-printed (default) — Optimized for printed text, ~250 MB
large-printed — Higher accuracy, ~400 MB
base-handwritten — Trained on handwritten text
large-handwritten — Large variant for handwriting

Limitation: TrOCR is designed for single lines of text, not full pages. Pair with a layout detector to crop regions before OCR.

Installation: Enabled via candle-ocr or full feature in Cargo.

Selection: Configure via backend options:

{
  "ocr": {
    "backend": "candle-trocr",
    "backend_options": {
      "variant": "base-printed"
    }
  }
}

Candle GLM-OCR

Multilingual vision-language model (0.9B) for full-page document parsing. Supports text, tables, formulas, and charts with region-aware layout dispatch.

Language	Code
English	`eng` / `en`
Chinese	`zho` / `zh`
Japanese	`jpn` / `ja`
Korean	`kor` / `ko`
French	`fra` / `fr`
German	`deu` / `de`
Spanish	`spa` / `es`
Italian	`ita` / `it`
Portuguese	`por` / `pt`
Russian	`rus` / `ru`
Arabic	`ara` / `ar`
Hindi	`hin` / `hi`
Thai	`tha` / `th`
Vietnamese	`vie` / `vi`

Note: GLM-OCR is trained on multilingual data and accepts any language code, falling back gracefully for unsupported languages.

Installation: Enabled via candle-ocr or full feature.

Selection: Use --ocr-backend candle-glm-ocr and optionally set --ocr-language from the list above.

Configuration:

{
  "ocr": {
    "backend": "candle-glm-ocr",
    "backend_options": {
      "task": "ocr",
      "device": "auto",
      "layout_mode": "whole_page"
    }
  }
}

Candle PaddleOCR-VL

Lightweight vision-language model for multilingual document parsing. Supports text, tables, formulas, and charts.

Language	Code
English	`eng` / `en`
Chinese	`zho` / `zh`
Japanese	`jpn` / `ja`
Korean	`kor` / `ko`
French	`fra` / `fr`
German	`deu` / `de`
Spanish	`spa` / `es`
Italian	`ita` / `it`
Portuguese	`por` / `pt`
Russian	`rus` / `ru`
Arabic	`ara` / `ar`
Hindi	`hin` / `hi`
Thai	`tha` / `th`
Vietnamese	`vie` / `vi`

Note: PaddleOCR-VL is trained on 109+ languages and accepts any language code.

Installation: Enabled via candle-ocr or full feature.

Selection: Use --ocr-backend candle-paddleocr-vl with optional --ocr-language.

Configuration: Requires a local model path:

{
  "ocr": {
    "backend": "candle-paddleocr-vl",
    "backend_options": {
      "task": "ocr",
      "device": "auto",
      "model_path": "/path/to/paddleocr-vl-model"
    }
  }
}

Candle DeepSeek-OCR

Vision-language model combining SAM vision encoder, ViT/Qwen2, CLIP, and language decoder. Supports multilingual document parsing.

Language	Code
English	`eng` / `en`
Chinese	`zho` / `zh`
Japanese	`jpn` / `ja`
Korean	`kor` / `ko`
French	`fra` / `fr`
German	`deu` / `de`
Spanish	`spa` / `es`
Italian	`ita` / `it`
Portuguese	`por` / `pt`
Russian	`rus` / `ru`
Arabic	`ara` / `ar`
Hindi	`hin` / `hi`
Thai	`tha` / `th`
Vietnamese	`vie` / `vi`

Note: DeepSeek-OCR is trained on multilingual data and accepts any language code.

Installation: Enabled via candle-ocr or full feature.

Selection: Use --ocr-backend candle-deepseek-ocr with optional --ocr-language.

Configuration: Requires a local model path:

{
  "ocr": {
    "backend": "candle-deepseek-ocr",
    "backend_options": {
      "device": "auto",
      "model_path": "/path/to/deepseek-ocr-model",
      "version": 2
    }
  }
}

Selecting Languages

Configuration Precedence

Language selection follows the configuration cascade (highest to lowest priority):

CLI flag: --ocr-language eng or --ocr-language eng+deu
Inline JSON config: --config-json '{"ocr": {"languages": ["eng", "deu"]}}'

Config file: xberg.toml or xberg.yaml:

[ocr]
backend = "tesseract"
languages = ["eng", "deu", "fra"]

Default: Uses backend’s default language behavior (usually English or auto-detect)

Code Format

Tesseract: ISO 639-3 three-letter codes (e.g., eng, deu, fra)
PaddleOCR: ISO 639-1 two-letter codes or script-specific variants (e.g., en, de, zh_hans)
Candle VLMs: Accept major language codes (e.g., eng, en, zho, zh)

Multiple Languages

Combine languages with + in CLI, or use arrays in config:

xberg extract --ocr-language eng+deu+fra file.pdf

[ocr]
languages = ["eng", "deu", "fra"]

{
  "ocr": {
    "languages": ["eng", "deu", "fra"]
  }
}

Fallback Strategy

If a requested language is unavailable:

Tesseract: Fails with an error if the tessdata pack is not installed
PaddleOCR: Falls back to the model’s training base or returns an error
Candle VLMs: Accept all language codes and attempt recognition (graceful degradation)

Install missing language packs for Tesseract via your OS package manager before OCR execution.

Language Detection

Use the language-detection feature to auto-detect document language, then pass the detected language to OCR:

from xberg import (
    ExtractInput,
    ExtractionConfig,
    LanguageDetectionConfig,
    OcrConfig,
    extract,
)

config = ExtractionConfig(
    language_detection=LanguageDetectionConfig(enabled=True),
    ocr=OcrConfig(backend="tesseract", language=["eng"]),
)

output = await extract(ExtractInput(kind="uri", uri="document.pdf"), config)
document = output.results[0]

print(document.detected_languages)

OCR Language Support

Backend Language Coverage

Tesseract

PaddleOCR

Sceptre

Candle TrOCR

Candle GLM-OCR

Candle PaddleOCR-VL

Candle DeepSeek-OCR

Selecting Languages

Configuration Precedence

Code Format

Multiple Languages

Fallback Strategy

Language Detection

See Also