Speech Language Model

A class of large neural models that processes both speech and text in a single end-to-end framework, integrating tasks — automatic speech recognition, spoken language understanding, dialogue, speech generation — that traditionally required separate modular systems. Examples include OpenAI Whisper, Meta SeamlessM4T, Google Gemini's audio tower, and various open-source models like Qwen-Audio. Speech language models are increasingly embedded in accessibility tools (voice assistants, meeting captioners, AI-powered AAC) and bring new fairness concerns: errors can appear in understanding or generation rather than transcription, and established fairness metrics (Word Error Rate) do not straightforwardly extend to their outputs.