Translation Engines & AI Models
Choosing and switching between models
What's the difference between the DeepL, Google, and OpenAI models?
Each engine has different strengths: DeepL is known for nuanced, natural phrasing in supported language pairs; Google offers broad language coverage and strong general-purpose accuracy; OpenAI's models excel at context-aware translation in conversational speech. Results can vary by language pair, so testing more than one engine for your primary use case is recommended.
How do I switch AI Voice Recognition and Translation models in PairaVoice?
Go to Settings > Voice Recognition Model or Translation Model and select your preferred provider. You can change this at any time, including between conversations.
Which model should I use for medical or clinical conversations?
There's no single "best" engine for every clinical scenario — accuracy varies by language pair and terminology. We recommend testing your most common language pairs across engines in batch mode and standardizing on whichever performs most reliably for your specialty and patient population.
Which model should I use for educational settings?
Similarly, test each engine against your school or district's most common family languages. Many education users find streaming mode with any engine sufficient for everyday communication, reserving batch mode for higher-stakes meetings like IEP conferences.
Does changing models affect translation accuracy or speed?
Yes. Engines differ in both processing speed and translation quality depending on language pair and sentence complexity. Switching models is a good troubleshooting step if you notice consistent accuracy issues with a particular language.