Qwen3-VL LoRA Caption

Workflow allows you to use a Qwen3 VL model (2B-Instruct is good enough) to create description based prompt of an image is sees as well as tag based. The two are combined in the end to create a high quality and varied prompt. You will need to install any missing nodes so you can use it properly. First run will also download automatically the Qwen3 model that you select (default: 2B-Instruct)