Amid the increasing demand for vocational English proficiency in China's textile industry, particularly in dyeing and finishing sectors, traditional teaching methods fail to address the operational and communicative needs of students. This study constructs a multimodal discourse-based teaching framework integrating visual images and hands-on videos to improve vocational English instruction. Utilising a design-based research paradigm combined with mixed methods, the study first identifies four core industry task scenarios-process debugging, equipment operation, quality control, and accident handling-through Delphi expert consultation. A three-dimensional matrix mapping “linguistic task-visual modality-video elements” was developed to guide multimodal resource construction. A 16-hour controlled experiment involving 86 students demonstrated that the experimental group receiving multimodal instruction significantly outperformed the control group in three dimensions: execution of operational instructions (+24.7 points), terminology reproduction (2.1× increase), and reduced cognitive load across six NASA-TLX indices. Qualitative data revealed mechanisms such as subtitle-gesture anchoring and visual warning guidance that supported terminology internalisation and error avoidance. Based on these findings, a vocational multimodal discourse production model was proposed, alongside a set of teaching resource development standards. This model provides a transferable framework for cross-disciplinary applications in vocational education. Future work should explore low-cost, scalable models for regions with limited access to VR and interactive platforms to expand the accessibility of multimodal English education.

multimodal discourse, visual images, hands-on videos, vocational English teaching, textile dyeing and finishing, design-based research, language task migration