ConvoZen.AI, NoBroker's enterprise conversational artificial intelligence agent platform, has entered the West Asian market by introducing two proprietary speech models—Alif and Rawi. These models are specifically designed for the Arabic language, reflecting how it is used in real life: bilingually, accounting for multiple dialects, and with active code-switching with English.
Functionality of the New Models
The Alif model is a bilingual speech-to-text model for Arabic and English. It is created to transcribe the complex reality of daily conversations in the Middle East and North Africa (MENA), which often involve mixing dialects and languages, unlike the pure, formal Arabic used to train most commercial Automatic Speech Recognition (ASR) systems.
The Rawi model handles the text-to-speech side. This is an Arabic text-to-speech model capable of imitating the natural rhythm, intonation, and flexibility of various dialects, which is necessary for business-to-consumer (B2C) interactions and in the corporate sector.
Development Experience and Implementation
Akhil Gupta, Co-founder and CTO of NoBroker, noted that experience across different markets has shown that people rarely speak in the pure, formal language on which most voice AI is trained. He emphasized that users often mix languages mid-sentence and use regional accents, requiring the system to be able to follow this style. He added that Alif and Rawi are not merely adaptations of an existing model for a new market but were built from scratch considering the actual pronunciation of Arabic.
This launch comes months after ConvoZen introduced its conversational AI stack along with advanced speech models in India. In India, the Akshara (speech-to-text) and Ragini (text-to-speech) models were developed to address the infrastructural challenges related to conversational interaction facing NoBroker. ConvoZen unifies the entire lifecycle into a single platform based on conversational AI agents, AI co-pilots, AI supervisors, and customer service agents.
Application in India
These two models are currently used by over 40 enterprises in nine Indian languages, including companies such as HDFC Bank, Tata AIG, Cars24, Apollo, and Maruti Suzuki. The Ragini model, which supports six languages, is adapted for use in the banking, financial services, and insurance (BFSI), automotive, healthcare, direct-to-consumer (D2C), and educational technology (edtech) sectors.