PP-OCRv5_server_rec
PaddlePaddle/PP-OCRv5_server_rec
A popular open image-to-text model, with 189.4K downloads a month. gigarouter benchmarks and hosts it as an OpenAI-compatible API.
about this model
PP-OCRv5_server_rec is an image-to-text (text line recognition) model that supports Simplified Chinese, Traditional Chinese, English, and Japanese, as well as complex text scenarios including handwriting, vertical text, pinyin, rare characters, rotation, distortion, and artistic text — all within a single model.
Key strengths
- Single model covering multiple languages and challenging conditions (handwriting, rotation, distortion, artistic text).
- Designed for high-accuracy line-level recognition; any character error marks the entire line as incorrect, ensuring practical reliability.
- Suitable for OCR pipelines requiring robust text recognition across diverse real-world inputs.
Benchmark accuracy
The following table shows per-scenario accuracy (line-level). A line is considered correct only if every character (including punctuation) is recognized correctly.
| Handwritten Chinese | Handwritten English | Printed Chinese | Printed English | Traditional Chinese | Ancient Text | Japanese | General Scenario | Pinyin | Rotation | Distortion | Artistic Text | Average |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0.5807 | 0.5806 | 0.9013 | 0.8679 | 0.7472 | 0.6039 | 0.7372 | 0.5946 | 0.8384 | 0.7435 | 0.9314 | 0.6397 | 0.8401 |
Example output
Text recognition result on a sample image (score 0.9535):
Pipeline usage
When combined with text detection and optional preprocessing modules, PP-OCRv5_server_rec can be used in a full OCR pipeline. Example output on a multi-line document:
We're benchmarking and onboarding PP-OCRv5_server_rec as a hosted, OpenAI-compatible API. Sign in for free credit and be ready when it lands, or tell us you want it and we'll prioritize it.