A neural network architecture designed for speech recognition that processes audio sequentially and outputs text in real-time without waiting for the entire input.