NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI · #2028 most downloaded on PyPI
Awesome OCR Library
Last release 2 months ago
21 Jul 2026
Ships fairly regularly
a new release about every 3 weeks
Nearly every release is documented
notes for 28 of 29 stable releases
2 versions withdrawn
withdrawn after publishing
2 years old
31 releases · first in 2025
One column per month.
add use_preprocess_img and use_vertical_padding logic by @SWHL in 3cc33f8
Full Changelog: 3.9.2
(python) add PP-OCRv6 PaddlePaddle backend support ( #696 ) by @JinmingMei in 29b812f
Full Changelog: 3.9.1
Supplemental Notes for Update Log
Full Changelog: v3.8.4...v3.9.0
Full Changelog: https://github.com/RapidAI/RapidOCR/compare/v3.8.4...v3.9.0
optimize the wheel packaging process (#687) by @SWHL in 86b48d2
rapidocr>=3.8.4)Full Changelog: 3.8.4
(python) include nested package resources in wheel by @SWHL in 2efe31f
add default download model configuration by @SWHL in a4c7382
Full Changelog: 3.8.2
add missing DEFAULT_MODEL_PATH property ( #664 ) by @SWHL in d4bf8f2
add RapidOCR Skill in the ClawHub
Full Changelog: 3.8.0
(inference) add get_character_list abstract method to InferSession by @SWHL in e549f7e
Full Changelog: 3.7.0
replace deprecated openvino.runtime import ( #626 ) by @moesnow in 4c39ce9
Full Changelog: 3.6.0
update five multi language files (#619) by @SWHL in 8487133
Full Changelog: 3.5.0
correct incorrect character coordinates in returned data (#614) by @SWHL in df8c571
Default parameters respect configuration file settings. (#591) by @hwf1324 in 3be5925
Full Changelog: 3.4.3
fixed issue #542 (#564) by @SWHL in 13fb8cc
Co-authored-by: yuege969 mingyue.zhou@senscape.com.cn by @yuege969 in eb6c685
Full Changelog: 3.4.2
fixed issue #554 by @SWHL in dec6056
remove deprecated getLevalName by @SWHL in 386c415
这三个语种默认对应LangRec.EN、LangRec.TH和LangRec.EL。具体可参见官方文档中模型列表一节。
下面给出希腊文识别示例：
from rapidocr import LangRec, ModelType, OCRVersion, RapidOCR
engine = RapidOCR(
params={
"Rec.lang_type": LangRec.EL,
"Rec.model_type": ModelType.MOBILE,
"Rec.ocr_version": OCRVersion.PPOCRV5,
}
)
img_url = "https://github.com/RapidAI/RapidOCR/blob/main/python/tests/test_files/el_rec.jpg?raw=true"
result = engine(img_url, use_det=False, use_cls=False, use_rec=True)
print(result)
result.vis("vis_result.jpg")
from rapidocr import RapidOCR
engine = RapidOCR(params={"Global.log_level": "critical"}) # 注意这里改为critical，就不会打印日志了
img_url = "https://img1.baidu.com/it/u=3619974146,1266987475&fm=253&fmt=auto&app=138&f=JPEG?w=500&h=516"
result = engine(img_url)
print(result)
result.vis("vis_result.jpg")
print(result.to_markdown())
都统一改为从前处理、模型推理和后处理都算上。
主要是拆分原有的test_main.py文件，利于后续维护
Full Changelog: 3.4.0
```python from rapidocr import RapidOCR
from rapidocr import RapidOCR
engine = RapidOCR()
img_url = "https://img1.baidu.com/it/u=3619974146,1266987475&fm=253&fmt=auto&app=138&f=JPEG?w=500&h=516"
result = engine(img_url, return_word_box=True, return_single_char_box=True)
print(result)
result.vis("vis_result.jpg")
print(result.to_json())
结果示例：
[{'box': [[71.0, 363.0], [419.0, 363.0], [419.0, 382.0], [71.0, 382.0]], 'txt': '曲曲折折的荷塘上面，弥望的是田田的叶', 'score': 0.99052}, {'box': [[71.0, 392.0], [410.0, 392.0], [410.0, 412.0], [71.0, 412.0]], 'txt': '子。叶子出水很高，像亭亭的舞女的裙。', 'score': 0.99123}, {'box': [[185.0, 466.0], [317.0, 466.0], [317.0, 488.0], [185.0, 488.0]], 'txt': '—《荷塘月色》一', 'score': 0.89048}]
from rapidocr import EngineType, LangDet, LangRec, ModelType, RapidOCR
from rapidocr.utils.typings import OCRVersion
engine = RapidOCR(
params={
"Rec.lang_type": LangRec.ESLAV, # KOREAN / LATIN / ESLAV
"Rec.engine_type": EngineType.ONNXRUNTIME,
"Rec.ocr_version": OCRVersion.PPOCRV5,
}
)
img_path = "tests/test_files/eslav.jpg"
result = engine(img_path, use_det=False, use_cls=False, use_rec=True)
print(result)
result.vis("vis_result.jpg")
Full Changelog: 3.3.0
采纳了小伙伴@qianliyx 的建议，按照行返回单字坐标：同一行的单字坐标是在同一个 tuple 中的。
from rapidocr import RapidOCR
engine = RapidOCR()
img_url = "https://img1.baidu.com/it/u=3619974146,1266987475&fm=253&fmt=auto&app=138&f=JPEG?w=500&h=516"
result = engine(img_url, return_word_box=True, return_single_char_box=True)
print(result)
result.vis("vis_result.jpg")
print(result.to_markdown())
Full Changelog: 3.2.0
(rapidocr) fixed typos of directml and npu by @SWHL in cec934a
adapt PR #451 by @SWHL in 41d5824
(rapidocr) integrated PP-OCRv4_dervr_rec-doc model (#426) by @SWHL in 80c72f5. See link for details.
(rapidocr) add whether is an interactivate window by @SWHL in 421e734
Full Changelog: 2.0.7
*(rapidocr)* Fix invalid classifier
*(rapidocr)* Sup more lang model for torch inference
*(rapidocr)* Support to other language ocr and update visualization (#369) by @SWHL in https://github.com/RapidAI/RapidOCR/pull/369
Nothing published for this version
*(rapidocr)* Support img url as input and optimize vis
(rapidocr) Support img url as input and optimize vis (#356)
rapidocr -img "https://github.com/RapidAI/RapidOCR/blob/main/python/tests/test_files/ch_en_num.jpg?raw=true" --vis_res
⚠️注意：该版本不与rapidocr_onnxruntime<=1.4.4兼容。遵循语义化版本号规则，请谨慎使用，以免出现各种问题。
⚠️注意：该版本不与rapidocr_onnxruntime<=1.4.4兼容。遵循语义化版本号规则，请谨慎使用，以免出现各种问题。
rapidocr库，整合onnxruntime、openvino、torch和paddlepaddle四个推理引擎，大家可按需安装对应的推理引擎来使用。rapidocr==2.0.1lang_det和lang_rec来实现。至此基本功能均已经实现。相关issue #329
Full Changelog: https://github.com/RapidAI/RapidOCR/compare/v2.0.0...v2.0.1
Nothing published for this version
Your coding agent can read these notes before it upgrades. Set up the MCP server →