Open Surebob opened 1 month ago
Hi @Surebob -- can you please describe the desired behavior vs. the behavior you're currently observing? Currently the app uses OpenAI's automatic server-side turn detection which is documented here
With translation between two humans + an AI, turn detection and interruption is actually a bit more nuanced when compared to 1 human + 1 AI.
If you think about the analog of having a human translator in a live conversation, often times Person A speaks their full utterance, the translator then translates once Person A is finished speaking, then Person B speaks followed by the translator translating back to Person A's language once Person B is done speaking. This behavior is how we modeled the app.
If you could give us a better sense of what you're looking for, perhaps that could help determine what's possible.
are there plans to implement a (correctly) working interruption mechanism?