Frontier voice-native models for transcription, deepfake detection, emotions, and behavior signals, available as REST and streaming APIs.