I’d like to suggest a small UX improvement regarding model switching between normal text chat and Voice mode.
When a conversation starts in a text model and Voice mode is later enabled in the same thread, the Voice/Live model takes over that conversation while Voice is active. That part is understandable.
What is less clear is what happens after Voice mode is turned off.
From a user perspective, there is no obvious indication of whether the conversation has returned to the previously active text model, remained on Live routing, or switched to something else. Typing again in the same thread does not make this clear.
This matters especially in long-running technical, coding, research, or project conversations where users may intentionally choose a specific text model because of its reasoning or analysis characteristics.
A possible improvement would be to make the active model state more transparent and give users explicit control when switching between Voice and text.
For example:
- clearly show which model is currently handling the conversation;
- after leaving Voice mode, automatically return to the previously active text model;
- or provide a visible “Return to previous text model” option;
- keep Voice mode and text-model selection independent where possible.
This is not necessarily a bug — there may be a valid internal reason for the current routing behavior. The main issue is that the transition is not transparent to the user.
Making the model state explicit would make Voice integration much more predictable, especially for users who maintain long-term project threads.