Cheng Cui

Tech Lead @ PaddleOCR | Document AI & VLM Researcher.

💡 Making AI practical and scalable.

[Cheng Cui @slimcat0101]

Appreciate the shoutout, Tiezhen! 🙏

The team is thrilled to see PP-OCRv6 holding the top 17 spots on the HF leaderboard. A rare milestone for us, and we couldn’t have done it without this incredible open-source community.

Keep the feedback coming! 🚀

[Tiezhen WANG @Xianbao_QIAN]

wow @huggingface Image-To-Text page has became @PaddlePaddle OCR v6 page. Is it so good?

[Cheng Cui @slimcat0101]

@ma_sc_ @PaddlePaddle Got it. A 100% reproduction rate on English and Italian is a strong signal for us. Thank you for such a solid batch of feedback. I'll have the team spin up a dedicated test using your flyer samples. Appreciate your help in making PP-OCRv6 better!

[Cheng Cui @slimcat0101]

@ma_sc_ @PaddlePaddle Any specific bad cases you could share?

[Cheng Cui @slimcat0101]

@pabis_eu @PaddlePaddle The models you mentioned are VLMs tailored for document OCR, rather than natural scene images. PP-OCRv6, on the other hand, is designed to handle text spotting in natural images. They belong to different technical paradigms.

[Cheng Cui @slimcat0101]

@ma_sc_ @PaddlePaddle Glad you like the model! I’m the tech lead for PaddleOCR. This word-merging issue is a known pain point. Which version are you currently on? We are making sure PP-OCRv6 addresses this spacing problem.

[Cheng Cui @slimcat0101]

Big update! 🎉 From now on, PaddleOCR inference is completely decoupled from the PaddlePaddle framework. You now have the absolute freedom to choose ONNXRuntime or PyTorch! 🚀

[PaddlePaddle @PaddlePaddle]

🚀 PaddleOCR 3.7 is here — ONNX Runtime now supported!

🔥 One parameter to switch inference backends. Zero code changes.

📊 What's new:
🔸ONNX Runtime backend: cross-platform, lightweight, multi-hardware acceleration.
🔸Multi-hardware support: CPU via OpenVINO, GPU via CUDA & TensorRT.
🔸PP-OCRv6 also onboard: Tiny / Small / Medium — up to 3.9× faster on CPU.

✨ From servers to edge devices — flexible deployment, your choice of engine.

🔗 Try it:
🌐 https://t.co/qf6cyafiqY
💻 https://t.co/oNOfB6hbSY
📖 https://t.co/yq12V9Ojhq

#PaddlePaddle #PaddleOCR #OCR #ONNX #ONNXRuntime

[Cheng Cui @slimcat0101]

@AI_Jasonyu Thanks everyone for your interest in PaddleOCR! I’m the tech lead of PaddleOCR. I haven't been very active on X in the past, but I'm planning to share more technical insights and updates here moving forward. Stay tuned! 🐼

[Cheng Cui @slimcat0101]

@servasyy_ai 感谢大家对 PaddleOCR 的关注！我是这个项目的技术负责人。平时确实很少上 X，之后打算在这里多分享一些技术干货和日常思考，欢迎大家多交流！🤝

[Cheng Cui @slimcat0101]

PP-OCRv6 released! Forget heavy VLMs for dedicated OCR. We've completely rewrote the rules of speed and efficiency!
97ms latency per image running locally in Web browsers! Absolute real-time response.
Drop a ⭐️ and clone it now! 👇 Repo Link: https://t.co/WEAlRNxEru

[Cheng Cui @slimcat0101]

🎉 80K stars - a HUGE milestone!
And guess what? PP-OCRv6 is coming! 🔥
⚡ Even FASTER speed
🎯 Even BETTER accuracy
Let's gooo! 🚀

[jzhang533 @jzhang533]

PaddleOCR just hit 80K STARS.

Kudos to the team! @slimcat0101 @liuyi_ai @PaddlePaddle

Recently we released PaddleOCR-VL-1.6, and is preparing the release of pp-ocr-v6. Looking ahead, we’re excited to push the boundaries of Document AI further.

🔗 https://t.co/J96qILm8cL

[Cheng Cui @slimcat0101]

Milestone! 🏆 PaddleOCR is now the #1 starred OCR repo on GitHub, surpassing Tesseract.

As the tech lead, I’m beyond thrilled! From a single line of code to a global choice—our mission remains "AI for everyone." Huge thanks to our 70k+ community. Let’s keep building! 🚀✨

[Cheng Cui @slimcat0101]

@PaddlePaddle @github @GitHubCommunity Let’s make document intelligence even better!😘

[Cheng Cui @slimcat0101]

@Zai_org Great work! I am Cheng Cui, the author of PaddleOCR-VL-1.5 and PP-DocLayoutV3. Thanks for using the PP-DocLayoutV3 model from PaddleOCR-VL-1.5. Let’s make document intelligence even better together!

[Cheng Cui @slimcat0101]

@uptonking2 @PaddlePaddle There can indeed be some challenges when using a two-stage model with frameworks like Ollama. Therefore, we offer a completely offline docker image for your convenience. Feel free to deploy it with a single click. Here is the documentation: https://t.co/wga8eaoFZq

[Cheng Cui @slimcat0101]

@getpochi @PaddlePaddle We will strive to ensure that the PP-DocLayout model in the first stage is as accurate as possible. Of course, the capabilities of this model are continuously being optimized.

[Cheng Cui @slimcat0101]

@CowPat101_ @PaddlePaddle What you said is absolutely right. We’re also considering how to make this part smoother.

[Cheng Cui @slimcat0101]

@NotEduardoP @PaddlePaddle Related work is underway

[Cheng Cui @slimcat0101]

@nicekate8888 感谢使用，相关信息我们会及时更新在github的PaddleOCR项目上！

[Cheng Cui @slimcat0101]

@rryssf Hey everyone, as the person in charge of this project, I saw this post and had to drop a comment! We’d love for you to give it a try. Your feedback is crucial—it’s what helps us make the tool better for everyone.😀