Not sure I understand what you're asking: there are faster than realtime models that you can use, locally or as APIs.