Discussions

I am currently using the Heygen library in my Angular project. In my workflow, when I initially call the avatar, I send the parameter repeat to the streaming.task endpoint. From that point onward, in order to prevent interruptions while the avatar is speaking, I need to mute the user’s microphone. To achieve this, I follow these steps:

avatar.startVoiceChat({ isInputAudioMuted: false });
avatar.muteInputAudio();

However, right after sending a request to the start endpoint, the avatar immediately detects the user’s voice, which sometimes causes the sentence I send through the task() function to be interrupted. Could you please advise on how I can prevent this behavior?

In addition, the text I send via task is being returned by Heygen in the user_talking_message as if it were spoken by the user. Furthermore, although we specify the language as tr and speak in Turkish, there are instances where Heygen returns the user’s speech in different languages or even different scripts. Since we log these conversations, this results in an inaccurate user experience.

Could you please clarify whether these issues are due to the method we are following on our side, or if they stem from something that needs to be resolved on your side?

I kindly request your support and guidance on this matter.

Thank you.