Awaken Open-Sourced: Turn a Half-Duplex Speech LLM Full-Duplex With Just a Small LoRA
alexcovo_eth · x · 2026-11-12
RealRichomie's team open-sources Awaken, a method that turns an existing half-duplex speech LLM into a full-duplex one that can listen and speak simultaneously by training just a small LoRA on top. Full-duplex voice models released recently sound amazing but require lots of data and money to build from scratch; with a small team and fixed budget, they found this cheaper path.
More from Multimodal
- A LoRA Is All You Need to Turn a Speech LLM Full-Duplex — kastnerkyle · 2026-11-12
- Runway teases triple bet: real-time avatars, interactive worlds, robot action models — c_valenzuelab · 2026-11-12
- MiniMax H3-Image Generates 4K Game Assets, Combined Into One Poster — azed_ai · 2026-11-12
- Simulating embroidery with Fable 5.2 and fal — OdinLovis · 2026-11-12
- Using fable 5.2 to compose a sonata played by gravity, no animation — imjustnewatai · 2026-11-12
- mark_k stress-tests new MiniMax H3 Image model with a dense 4K prompt — mark_k · 2026-11-12