Alibaba Introduces Fun-ASR1.5: Advancing Multi-language Speech Recognition

Alibaba has unveiled Fun-ASR1.5, a major upgrade to its end-to-end speech recognition model.

Fun_ASR1_5

Alibaba has unveiled Fun-ASR1.5, a major upgrade to its end-to-end speech recognition model. Supporting 30 languages—including Chinese, English, Japanese, Thai, French, and Arabic—the model ensures high-precision recognition across diverse linguistic regions globally.

A core strength of Fun-ASR 1.5 is its seamless code-switching capability. By utilizing a unified training framework, the model automatically detects and transitions between languages in mixed-lingual dialogues without manual configuration. This makes it an effective solution for multinational collaboration, multilingual content creation, and international conferences. In addition, for Chinese outputs, it delivers “ready-to-use” text featuring intelligent punctuation and automated formatting for dates, currencies, and numerals.

Fun-ASR1.5 is powered by a Mixture-of-Experts (MoE) architecture, which employs “on-demand” parameter activation to balance massive scale with high computational efficiency. Combined with a multi-stage pre-training strategy, the accuracy of the model has been enhanced.

Fun-ASR 1.5 is now available via API services on the Alibaba Cloud Model Studio platform. Users can also test out the features on ModelScope.

This article was originally published on Alizila written by Shao Xiaoyi and Claire Mo.

0 0 0

Share on

Community

Alibaba Introduces Fun-ASR1.5: Advancing Multi-language Speech Recognition

Read previous post:

Alibaba Cloud Community

You may also like

Comments

Alibaba Cloud Community

Related Products

Alibaba Cloud Model Studio

Tongyi Qianwen (Qwen)

Alibaba Cloud for Generative AI

AI Acceleration Solution