Japanese SFT/DPO data convert to speech via TTS. And audio caption data generated by Qwen3-Omni. All datasets are available for commercial use.