Info

• Title: LLaMA-Omni 2: LLM-based Real-time Spoken Chatbot with Autoregressive Streaming Speech Synthesis
• Group: CAS
• Keywords: Speech LLM, streaming
• Venue: Submitted to EMNLP 2025

Comments

1. Read and write: read and fuse from the text token, then generate the speech tokens corresponding to the text token.
2. Use Qwen2.5-0.5B for model initialization.