On September 20, at the Conference on High-Quality Development of Teacher Education and the 70th Anniversary of the Founding of Qinghai Normal University, Qinghai Normal University publicly released the "Zhida AI Input Method" and its companion Tibetan font library.

It is reported that this is China's first Tibetan multilingual omni-modal AIinput method. The input method was developed by the National Key Laboratory of Tibetan Intelligent Information Processing. Based on the laboratory's proprietary large model and related AI infrastructure, it covers all terminals, including mobile phones and computers, supports multimodal input such as text, speech, and images, and can sync the lexicon and input habits via a unified account.

There is a considerable gap between the spoken dialects and the written form of Tibetan, and mixed Chinese-Tibetan usage is relatively common. Dola, executive deputy director of the National Key Laboratory of Tibetan Intelligent Information Processing and professor at Qinghai Normal University, introduced that the "Zhida AI Input Method" integrates a Tibetan keyboard, Latin transliteration, mixed Chinese-Tibetan-English speech, image recognition, and AI assistance into a unified interface. Among these functions, speech input supports both written and spoken forms of the three major dialects, U-Tsang (Central Tibetan), Amdo (Northeastern Tibetan), and Kham (Eastern Tibetan), and also supports mixed Chinese-Tibetan-English speech. The image recognition function can recognize mixed Chinese-Tibetan-English text from photos or screenshots, serving scenarios such as document digitization, teaching, and office work.

In terms of intelligent assistance, the input method supports automatic error correction, text polishing, and bidirectional translation, and offers multiple language styles, including formal, friendly, and academic. Its lexicon covers 23 disciplines, including education, law, and literature, and it can also learn and remember users' frequently used vocabulary and expression habits. Its technical architecture combines on-device and cloud computing and supports local inference, improving user experience while safeguarding user privacy.

The Zhida Tibetan font set, released alongside the input method, was developed in accordance with international standards. It includes Hei (sans-serif), Bai (white/regular), artistic, and brush styles, and is optimized for font size adaptation in mixed Chinese-Tibetan typesetting. It can be used for screen display, printing and publishing, and artistic design.

The Zhida large model was also showcased at the launch event. The model comes in nine versions, with parameter sizes ranging from 800 million to 100 billion. Three main models are deployed online, supporting mutual translation among 56 languages as well as image and video understanding, and can serve fields such as education, scientific research, government affairs and so on.

Reporter: Sun Rui

Translator: Ni Weijing

First Review: Zhi Xinghua

Second Review: Drakpa Wangchen

Third Review: Hu Rongguo