PaddlePaddle/PaddleOCR · error · ValueError

No models are available for lang={lang} and ocr_version={ocr

Error message

No models are available for lang={lang} and ocr_version={ocr_version}.

What it means

When no explicit detection/recognition model names or dirs are given, PaddleOCR looks up default model names for the (lang, ocr_version) pair via _get_ocr_model_names; if either lookup returns None it raises ValueError. This means the language code is unknown or has no models for that OCR version — it is a table-lookup failure, not a network failure.

Source

Thrown at paddleocr/_pipelines/ocr.py:117

                f"Invalid OCR version: {ocr_version}. Supported values are {_SUPPORTED_OCR_VERSIONS}."
            )

        if all(
            map(
                lambda p: p is None,
                (
                    text_detection_model_name,
                    text_detection_model_dir,
                    text_recognition_model_name,
                    text_recognition_model_dir,
                ),
            )
        ):
            det_model_name, rec_model_name = self._get_ocr_model_names(
                lang, ocr_version
            )
            if det_model_name is None or rec_model_name is None:
                raise ValueError(
                    f"No models are available for lang={repr(lang)} and ocr_version={repr(ocr_version)}."
                )
            text_detection_model_name = det_model_name
            text_recognition_model_name = rec_model_name
        else:
            if lang is not None or ocr_version is not None:
                warnings.warn(
                    "`lang` and `ocr_version` will be ignored when model names or model directories are not `None`.",
                    stacklevel=2,
                )

        params = {
            "doc_orientation_classify_model_name": doc_orientation_classify_model_name,
            "doc_orientation_classify_model_dir": doc_orientation_classify_model_dir,
            "doc_unwarping_model_name": doc_unwarping_model_name,
            "doc_unwarping_model_dir": doc_unwarping_model_dir,
            "text_detection_model_name": text_detection_model_name,
            "text_detection_model_dir": text_detection_model_dir,

View on GitHub (pinned to 2661c7c0ef)

Solutions

Use a supported language code, e.g. lang='ch' or lang='en' — check the docs' language list for the release.
Try a different ocr_version (e.g. PP-OCRv4 or PP-OCRv5) that has models for that language.
Bypass the lookup entirely by specifying explicit models: text_detection_model_name and text_recognition_model_name (or local model dirs).

Example fix

# before
ocr = PaddleOCR(lang='chinese', ocr_version='PP-OCRv5')  # ValueError
# after
ocr = PaddleOCR(lang='ch', ocr_version='PP-OCRv5')
# or pin explicit models
ocr = PaddleOCR(text_detection_model_name='PP-OCRv5_server_det',
                text_recognition_model_name='PP-OCRv5_server_rec')

Defensive patterns

Strategy: validation

Validate before calling

from paddleocr._pipelines.ocr import PaddleOCR

def has_models_for(lang, ocr_version) -> bool:
    det, rec = PaddleOCR._get_ocr_model_names(lang, ocr_version)
    return det is not None and rec is not None

Prevention

Use documented short language codes ('ch', 'en', 'japan'), never full names
Test (lang, ocr_version) pairs programmatically at config-load time
Fall back to explicit model names when the default table lacks coverage

When it happens

Trigger: PaddleOCR(lang='xx', ocr_version='PP-OCRv5') with an unsupported language code; combining a valid lang with a version that lacks models for it (e.g. a lang not covered by PP-OCRv6); typo'd lang like lang='chinese' instead of 'ch'.

Common situations: Using full language names ('english', 'chinese') instead of codes ('en', 'ch'); copying lang lists from other OCR libs; requesting the newest OCR version for a language whose models only exist for older versions; passing None lang with an ocr_version for which defaults are unavailable.

Related errors

AI-assisted analysis of PaddlePaddle/PaddleOCR@2661c7c0ef (2026-08-14). Data as JSON: /api/errors/1d821a338ebc8427. Report an issue: GitHub.