Discussions

Hey Heygen team and community,

Working on something that would require the interactive avatar to understand and speak in Arabic, and hoping to get some suggestions as to how to best implement something like this.

When running createStartAvatar with language set to "ar", i get a response stating that Deepgram doesn't support Arabic language STT. So the next step for us is to have a separate STT provider.

But this seems to bring on multiple problems while trying to maintain the interactivity and quick conversational nature of the avatar.

• Like its ability to respond quickly, since a whole new request would have to go to the STT provider, which will have to respond with the Arabic text derived from the speech, which will then be sent to the avatar. This would add a significant delay to a response from the avatar.
• Also using a button to start and stop recording like mentioned in the docs here, which makes it harder to ensure a conversational nature for our use case.
• I've also been recommended to play around with these flags and methods provided: interrupt(), startListening(), stopListening(), USER_START, USER_STOP, USER_SILENCE. I'm hoping someone can elaborate on how to use these for some better success.

If anybody has dealt with a similar problem before, please let me know how you've tried to resolve it and what similar steps I can take to make this use case work out.

If anybody has suggestions to how I can change up the architecture flow of the app or if I've made a mistake somewhere in my explanation of the problem, please let me know.

Again, my main difficulty in all of this is maintaining interactivity and conversations as similar to a human conversation as possible.

Thanks,
Mohit.