There is good news about language technology. Users will now be able to translate the text in their voice. Technology company Google has recently unveiled its new language model AudioPaLM. Developed by the researchers of Google, this platform will provide a new facility to the users. This language model can perform well in listening, speaking and translating. According to techlusive.in news, AudioPaLM is a multimodal architecture that combines the advantages of two existing models – PaLM-2 and AudioLM.
How does this model work
According to the news, PaLM-2 is a text-based language model that is efficient in understanding text-based special linguistic knowledge. AudioLM is adept at maintaining information such as speaker identity and tone. By combining these two models, AudioPaLM uses the linguistic capabilities of PaLM-2 and the lexical information preservation of AudioLM to enable deeper understanding and construction of both text and speech.
Ability to transfer voice to multiple languages
The language model AudioPaLM can also perform zero-shot speech-to-text translations for multiple languages, even for speech combinations it didn’t see during training. This capability can be useful for real-world applications such as real-time multilingual communication. AudioPaLM can transcribe voices in different languages based on short spoken signals, and it can capture and reproduce different voices in different languages. Can do.
