Local large models typically refer to open-source LLMs (such as Llama and Qwen series) hosted by tools like Ollama and LM Studio, responsible for text-side tasks such as script generation, subtitle writing, and intelligent scheduling