An app that helps you learn Chinese by reading with the help of AI. Depending on a system, it might require to install ONNX Runtime.
OCR uses PaddleOCR, a Chinese engine that is superior to Tesseract and other popular OCR engines in recognizing Chinese characters.
Download the following files from Hugging Face
- https://huggingface.co/monkt/paddleocr-onnx/blob/main/detection/v5/det.onnx
- https://huggingface.co/monkt/paddleocr-onnx/blob/main/languages/chinese/rec.onnx
- https://huggingface.co/monkt/paddleocr-onnx/blob/main/languages/chinese/dict.txt
And put them inside the models folder in your HOME directory (C:\Users{user name}\models or /home/{user name}/models)
Either use the image button to load image from clipboard, or load an image file.
Either subscribe to one of the supported providers to get an API key or set up a local Ollama chat. Among local models, Gemma4 is recommended for this task.
The functions available:
- examples: give a few examples of the usage of selected word
- meaning: explain meaning without context
- grammar: explain grammar of a selected fragment
- explain: explain the meaning of a given fragment, with the whole text in context
- translate: to your target language (app language)
It allows you to save notes so you can go back to them and review later. Can also add the dictionary or AI output to notes.
Here pick the third button to manage texts in a database.
Here you can:
- select one of the texts in the base and click the first button to load it
- delete selected text
- add a new text (it will ask you to provide the name), saving the contents that are in the main window
- cancel
If you use local AI models with llama.cpp, you can manage configurations and start/stop llama-server from this app now.
Just show the paths to model and mmproj (if you want vision) and then use the button with coloured square:
- ⬜️ - local llama.cpp found, not running
- 🟩 - running, ready
- 🟥 - starting, not ready, not responding
- 🟦 - running, ready, but not started from this app
- 🟪 - streaming response
- ⬛️ - not available For Ollama users to the left of it there is a "toggle keep alive" button so it does not unload a model while reading.
The other buttons (left to right)
- copy AI response to the text tab
- copy response to notes
- toggle Ollama keep alive
- llama.cpp state
- cancel currently running AI prompt
- clear the conversation history
Improved prompter lets you choose whether to include the text and/or a loaded image with your prompt.
If you write empty brackets {} in the chat, it will replace them with the selected fragment of text, e.g.
Explain the meaning of the following fragment: {}
Here you choose the language and set up your API key.