Artificial Intelligence-Driven Interactive Experience for Intangible Cultural Heritage: Sustainable Innovation of Blue Clamp-Resist Dyeing ^†

Abstract

The sustainable development of intangible cultural heritage (ICH) faces multiple challenges, including societal structural transformations, intergenerational transmission gaps, and the loss of cultural memory. The rapid advancement of generative artificial intelligence (AI) technologies offers new possibilities for the digital preservation and innovation of ICH. This study leverages generative AI to develop a LoRA model embodying the Blue Clamp-Resist Dyeing style, enabling the digital preservation and innovative reinterpretation of traditional patterns. Additionally, the study integrates these technological achievements into an interactive experience project at the Wenzhou Blue Clamp-Resist Dyeing Museum. Through immersive experiences in pattern generation and dissemination, the project effectively enhances public engagement and cultural identity. The findings reveal that generative AI holds significant potential for promoting the digital transformation, innovative dissemination, and sustainable development of ICH. This study offers a practical approach to the preservation and innovation of intangible cultural heritage. By applying generative artificial intelligence, it further expands the potential for enhancing public engagement and promoting innovative cultural heritage transmission. Additionally, it provides new possibilities for leveraging digital technologies to support the sustainable development of intangible cultural heritage.

1. Introduction

The protection, transmission, and utilization of intangible cultural heritage (ICH) have become pressing global concerns, drawing significant academic and practical attention. In China, much ICH is deeply embedded in rural communities [1]. However, shifts in social structures, generational gaps in transmission, and the erosion of cultural memory threaten the sustainability of ICH [2]. This creates a dual challenge; safeguarding the cultural integrity of ICH while addressing the imperatives of sustainability and innovation in contemporary society [3].

The transmission of ICH is often deeply rooted in specific cultural contexts and practical settings. Generative artificial intelligence technology, which can transform text into high-quality visual imagery [4], overcomes the constraints of time and space. This enables traditional cultural symbols to be digitally visualized [5], offering a more intuitive way to present and disseminate ICH [6].

The application of AI technology in the digital preservation of ICH, diversification of dissemination methods, and design innovation has opened new avenues for its sustainable development. Q. Liu et al. investigated the significant role of generative AI in the design of marine folk culture on the Jiaodong Peninsula, achieving the digital reconstruction and reproduction of traditional aesthetic elements [7]. U. C. Ajuzieogu’s research emphasized the role of generative AI in ICH reconstruction, proposing the use of AI technology to digitize archives and employ generative models for preserving ICH content such as music and handicrafts [8]. K. Nie et al. integrated AR and VR technologies to create immersive experiences for Chinese kite culture, demonstrating how AI enhances the interactivity and entertainment value of ICH content [9]. Similarly, K. H. C. Lau et al. employed VR and eye-tracking technologies to study user engagement in learning the cultural practices of Neapolitan pizza-making, further showcasing the potential of personalized AI technologies in ICH education [10]. C. Binucci et al. designed an intelligent recommendation system called CHIP for cultural tourism route planning [11]. Based on user preferences and cultural backgrounds, this system provides personalized ICH experience recommendations, exemplifying the integration of ICH with modern technology. S. Gola et al. expanded this research field by combining temporal planning and knowledge representation techniques to develop a personalized tourism route generation system, enhancing the dissemination of ICH [12]. Despite the significant role AI plays in ICH preservation, its complexity remains a barrier to widespread adoption. M. Dai et al. highlighted that technological complexity limits public participation in ICH initiatives and recommended designing user-friendly interfaces to improve accessibility and engagement [13]. These approaches contribute to the revitalization and sustainable development of ICH [14].

In conclusion, artificial intelligence has provided significant technical support for the preservation, transmission, and innovative design of intangible cultural heritage. However, its practical application and implementation in modern contexts remain critical focal points for future research. Additionally, generative artificial intelligence technology can dynamically document cultural changes, offering technological support for the “living” transmission of ICH, and helping to bridge the divide between tradition and modernity, as well as aesthetics and sustainability.

Blue Clamp-Resist Dyeing, a distinctive branch of the tie-dyeing technique, represents one of the key intangible cultural heritages of the Wenzhou region in China. This traditional Chinese handcraft dyeing method, with a history spanning over a thousand years, is an integral part of China’s cultural heritage. Renowned for its unique patterns, artisanal craftsmanship, and profound cultural significance, Blue Clamp-Resist Dyeing has become a symbol of Chinese folk art. However, the sustainable development of this craft faces substantial challenges due to the influence of foreign cultures, changes in production techniques, shifts in public aesthetic preferences, and limited market demand.

Generative artificial intelligence technology presents a promising opportunity for the sustainable development of this traditional craft [15]. By employing intelligent algorithms, generative artificial intelligence technology can generate digital representations of Blue Clamp-Resist Dyeing, offering innovative approaches to its preservation and expanding its modern applications, thereby revitalizing it in the digital age [16]. Through generative artificial intelligence technology, the cultural elements of Blue Clamp-Resist Dyeing can be digitally recreated in diverse cultural contexts, such as intangible cultural heritage museums, allowing a wider audience to explore and appreciate this traditional technique. Additionally, design concepts produced by generative artificial intelligence technology can be applied to various fields, such as fashion and contemporary art, facilitating the integration of Blue Clamp-Resist Dyeing into modern design trends and opening up new market opportunities.

Existing studies have systematically examined the craftsmanship and cultural significance of blue calico from the perspective of intangible cultural heritage preservation. However, there is a notable research gap in its integration with modern technologies, particularly generative artificial intelligence. Practical research on the sustainable innovation and development of blue calico within the context of digitalization remains limited. This study aims to explore how artificial intelligence can be employed to promote the sustainable development and digital innovation of blue calico, offering new perspectives and pathways for the preservation and innovation of intangible cultural heritage.

Based on the research objectives, this paper will explore the following three research questions:

RQ1: How can the cultural logic and integrity of Blue Clamp-Resist Dyeing patterns be ensured?

RQ2: How can generative artificial intelligence be leveraged to achieve the digital innovation of Blue Clamp-Resist Dyeing?

RQ3: How can generative artificial intelligence be utilized to enhance interactive experiences and cultural dissemination in intangible cultural heritage museums?

2. Literature Review

2.1. Intangible Cultural Heritage and Sustainable Development

In 1987, the United Nations World Commission on Environment and Development (WCED) defined sustainable development in its report Our Common Future as “a harmonious process of change that enhances the potential of both current and future generations to meet human needs and aspirations”. Since then, sustainable development has become a formal concept on the global agenda [17]. It has evolved from a focus on natural resource sustainability to a broader framework encompassing economic, environmental, social, and cultural dimensions, all of which are human-centered and socioeconomically driven [18]. Advocates of “cultural sustainability” have long emphasized the need to strengthen the relationship between culture and development for the sustainability of local communities [19], with this perspective emerging in the 1990s. Cultural diversity links intangible cultural heritage to sustainable development, with intangible cultural heritage serving both as a primary source of cultural diversity and a safeguard for sustainability [20]. As stated in Article 6 of the Action Plan for Promoting Cultural Policies for Development, adopted by UNESCO’s intergovernmental meeting on cultural policies, all member states unanimously agreed that “cultural creativity is the source of human progress; and cultural diversity, as a treasure of humanity, is an indispensable element for development” [21].

Therefore, as a key component of cultural diversity, intangible cultural heritage is vital for promoting sustainable societal development [22]. This paper explores the practical role of Blue Clamp-Resist Dyeing, a traditional form of intangible cultural heritage, in fostering cultural diversity and examines how it can be leveraged as a cultural resource to drive sustainable development. By integrating this heritage with generative artificial intelligence technology, the study aims to provide new insights into the sustainable development of intangible cultural heritage.

2.2. Generative Artificial Intelligence Technology

Generative artificial intelligence technology, based on fundamental artificial intelligence technologies such as machine learning, deep learning, and large model pre-training [23], enables the generation of digital content in various forms, including text, images, audio, and video. Recent advancements in generative artificial intelligence technology have led to significant breakthroughs in simulating human creativity and imagination [24], opening new possibilities in the fields of creation and design [25].

Currently, generative artificial intelligence technology is widely used in the design and creative industries [26], providing designers and artists with tools for creative presentation, generation, and expansion. By leveraging Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN), generative artificial intelligence technology can produce images in specific styles. Through training on large-scale datasets, generative artificial intelligence technology learns from diverse content, resulting in highly creative and unique design works [27].

In cultural institutions, particularly museums, the application of generative artificial intelligence technology has further expanded the potential of interactive design, offering innovative ways to enhance visitor experiences. In terms of smart navigation and interpretation, AI-powered digital avatars can engage with visitors through movements, facial expressions, and voice, delivering personalized services [28]. Additionally, personalized tour recommendation systems can tailor visit routes and exhibition suggestions based on individual interests and preferences [29]. In exhibit displays and immersive experiences, generative artificial intelligence technology is extensively applied in Augmented Reality (AR) technologies, allowing visitors to interact in a blended reality environment [30]. By incorporating generative AI technologies, museum narrative design has been enriched, significantly enhancing innovation and visitor engagement [31].

2.3. Blue Clamp-Resist Dyeing

Blue Clamp-Resist Dyeing is a form of blue-patterned fabric production and a remnant of historical multicolored clamp-resist dyeing techniques.

The production of Blue Clamp-Resist Dyeing involves floral pattern carving, indigo dye preparation, and the printing and dyeing processes. The workflow is illustrated in Figure 1. Floral pattern carving uses high-quality wood to create resist templates by engraving designs. Indigo dye is primarily extracted from plants such as Polygonum tinctorium or Isatis tinctoria, whose leaves are rich in indican. Through fermentation and natural oxidation, the indican is converted into indigo, which is further processed into powders or pastes for dyeing. During the printing and dyeing process, fabric is placed beneath the template, resist paste is scraped over it, and the material is repeatedly dipped into an indigo solution. The color gradually deepens with each immersion. After washing off the resist paste, a blue-and-white pattern emerges, showcasing the intricacy of the traditional craftsmanship and the eco-friendly nature of natural dyes [32]. The molecular structure and chromophore of indigo are shown in Figure 2.

According to the study by Z. X. Ju, the natural indigo dye used in Blue Clamp-Resist Dyeing exhibits excellent physicochemical properties. Its molecular structure is symmetrical and highly crystalline, containing key amide groups and aromatic rings. These characteristics confer good adsorption and stability during fiber dyeing. Spectral studies indicate that the main absorption peaks of indigo between 268 nm and 612 nm are attributed to π → π* transitions. Additionally, the fluorescence emission peak at 486 nm also corresponds to π → π* transitions. Compared with synthetic indigo, natural indigo has a more uniform molecular distribution, resulting in superior dyeing performance and certain antibacterial properties. These features provide Blue Clamp-Resist Dyeing with enhanced color fastness [33]. The infrared and Raman spectra are shown in Figure 3.

The patterns of Blue Clamp-Resist Dyeing are primarily linear, incorporating a combination of points, lines, and planes. These motifs often focus on human figures, drawn from folk legends, operatic stories, or everyday life scenes. Reflecting the folk life and historical development of the people in southern Zhejiang, these patterns express strong connections to their time, embodying the public’s aspirations and longing for beauty. Historically, Blue Clamp-Resist Dyeing was used as a bedding fabric for brides in the Wenzhou region. However, with the rise of machine-printed fabrics in the mid-20th century, the practice gradually faded from everyday use. Despite this, the patterns of Blue Clamp-Resist Dyeing continue to exhibit strong cultural characteristics and significant artistic value.

The traditional production process of Blue Clamp-Resist Dyeing has long relied on manual craftsmanship, following the apprenticeship system for knowledge transmission, which requires considerable time and physical labor. Over time, the aesthetic appeal of Blue Clamp-Resist Dyeing patterns has faced substantial challenges. Traditional designs, such as the “Hundred Sons” pattern, operatic character motifs, and other auspicious symbols, no longer align with contemporary tastes and preferences. The development of traditional woodblock printing and dyeing as an intangible cultural heritage is closely tied to the innovation of woodblock patterns. Therefore, researching the reinvention of Blue Clamp-Resist Dyeing patterns is crucial for providing insights and guidance on the continued evolution of traditional woodblock printing and dyeing techniques.

3. Materials and Methods

3.1. Research Process Design

This study has developed a research process based on the research content, as shown in Figure 4.

3.1.1. Theoretical Research Stage—Principles of Generative Artificial Intelligence Technology and Challenges in the Development of Blue Clamp-Resist Dyeing

To overcome the obstacles in the inheritance process of Blue Clamp-Resist Dyeing, this stage involves a systematic literature review and data compilation to study the aesthetic characteristics of its patterns. The research will explore the relationship between the aesthetics and cultural significance of Blue Clamp-Resist Dyeing patterns. Based on this, a detailed anatomical and systematic analysis of the patterns will be conducted, leading to the creation of a digital database. Furthermore, by applying digital recognition and coding to the patterns, this stage will provide the theoretical foundations and data support for subsequent experimental studies, laying the groundwork for the digital inheritance and innovation of Blue Clamp-Resist Dyeing.

3.1.2. Digital Preservation and Collection Stage—Building a Database of Blue Clamp-Resist Dyeing Patterns

This stage focuses on reviewing the principles of generative artificial intelligence technology, its current advancements, and its applications in the preservation of intangible cultural heritage and museum contexts. The objective is to establish the theoretical foundation and a knowledge map for the study. Through field surveys and a comprehensive literature review, the research will analyze the challenges faced by the traditional woodblock printing and dyeing technique of Blue Clamp-Resist Dyeing in its contemporary development. It will also explore how generative artificial intelligence technology can address the developmental bottlenecks of this traditional craft, uncover its cultural and aesthetic values, and inject innovative potential into its future evolution.

3.1.3. Digital Innovation Stage—Training the LoRA Model for Blue Clamp-Resist Dyeing Patterns

Building on the previously constructed pattern database, this stage applies deep learning and machine learning techniques to train the LoRA (Low-Rank Adaptation) model for generating Blue Clamp-Resist Dyeing pattern styles. This is the core phase of the study, focusing on the development of innovative pattern generation techniques that integrate contemporary aesthetic demands with the cultural context of Blue Clamp-Resist Dyeing. The research will prioritize both the artistic and cultural innovation of the patterns, aiming to facilitate the transmission and sustainable development of traditional woodblock printing and dyeing techniques in the modern era.

3.1.4. Digital Application Stage—Interactive Experience Design for the Blue Clamp-Resist Dyeing Museum

The ultimate goal of this study is to promote the sustainable development of intangible cultural heritage and facilitate cultural transmission and innovation through public participation. Guided by the principles of sustainable design, this stage proposes methods for reducing the carbon footprint in the inheritance and development processes [34], while using generative artificial intelligence technology to encourage deep public involvement in the creation of intangible heritage. Based on the completed LoRA model for Blue Clamp-Resist Dyeing, interactive applications for museums will be developed, allowing users to experience and engage in the innovative design of Blue Clamp-Resist Dyeing patterns. This will contribute to the learning, dissemination, and sustainable development of this intangible cultural heritage.

3.2. Research Methodology

Generative artificial intelligence technology has emerged as a prominent research topic in the field of artificial intelligence in recent years [35], offering the potential to replace human involvement in content generation at a low cost and on a large scale. AI-generated art, a subfield within generative AI, is particularly noteworthy [36]. Currently, mainstream AI art models, such as DALL·E 3, Imagen 3, Midjourney V6.1, Disco Diffusion v5.4.0, and Stable Diffusion v2.0, are capable of generating images quickly from natural language prompts or through image-to-image transformations [37]. Each of these models has its own strengths in the text-to-image generation domain, depending on their underlying training mechanisms. However, a common challenge shared by these models is the high variability in style and a lack of precise control over the generated images.

Historically, Blue Clamp-Resist Dyeing was primarily used as bedding fabric. Due to its delicate nature and the local tradition of including garments as burial items, well-preserved examples of Blue Clamp-Resist Dyeing are rare. The fabric’s fragility, combined with inadequate preservation conditions, presents challenges when attempting to apply generative artificial intelligence technology in this context.

To address these challenges, this study employs the Low-Rank Adaptation (LoRA) model, primarily due to its unique advantages in efficiently fine-tuning large pre-trained models, which align well with the needs of this research.

LoRA introduces low-rank matrices to efficiently adjust the parameters of pre-trained models. These enable the model to quickly adapt to new tasks while freezing the original weights, achieving this with minimal computational and storage costs [38]. This advantage is particularly suited for research scenarios that require an efficient and resource-friendly fine-tuning process. In this study, due to the limitations of traditional Blue Clamp-Resist Dyeing crafts in physical preservation environments, the quality and diversity of preserved patterns are constrained. Therefore, the research needs to achieve efficient adaptation with a small amount of data, making LoRA an ideal choice for this few-shot learning scenario. Moreover, the generation of intangible cultural heritage patterns requires the model to accurately capture complex cultural features and traditional styles. LoRA, by adjusting the parameters of a pre-trained large model, endows the model with new generative capabilities, ensuring both the diversity of generated patterns and a reduction in the computational burden, making it suitable for resource-constrained environments [39].

Compared with other AI models, LoRA has distinct advantages. First, unlike traditional full-model fine-tuning methods, LoRA adjusts model parameters through low-rank matrices, updating only a portion of the weights, significantly reducing computational and storage costs. Second, LoRA performs exceptionally well in few-shot learning, efficiently adapting to new tasks under limited data conditions and avoiding overfitting. In contrast, Textual Inversion only optimizes the text embedding space, offering less flexibility, while Hypernetworks have higher computational complexity and greater storage demands, and DreamBooth requires adjusting the weights of multiple modules, resulting in higher computational costs. With its concise and efficient low-rank adjustment method, LoRA is better suited for diverse generative tasks and is more appropriate for resource-limited environments [40].

4. Experimental Process and Results

4.1. Data Collection and Cleaning of Blue Clamp-Resist Dyeing Patterns

The digitization of patterns involves several key steps, including data collection, organization, recognition, and classification.

In the data collection phase of this study, two rounds of data acquisition were conducted. The first round involved capturing 57 images on-site at the Blue Clamp-Resist Dyeing Museum, while the second round entailed scanning 139 images from various books, including Wenzhou Cangnan Clamp-Resist Dyeing, Kunqu Opera on Blue Cloth, and The Master of Blue Clamp-Resist Dyeing Craft—Zhang Qin. In total, 196 images were collected across both rounds.

To ensure the authenticity and reliability of the model training results, the collected data were further refined. As shown in Figure 5, 132 high-quality Blue Clamp-Resist Dyeing patterns were selected as the core dataset for model training.

Building on these images, this study applies semiotic and semantic theories to deconstruct and classify the patterns based on thematic expression, morphological characteristics, and compositional structure. This classification process serves as the foundation for establishing a digital pattern database.

4.2. Analysis of Blue Clamp-Resist Dyeing Patterns

In terms of structure, Blue Clamp-Resist Dyeing patterns primarily consist of a main graphic, auxiliary graphics, and divider line, as illustrated in Figure 6.

Content-wise, Blue Clamp-Resist Dyeing patterns can be classified into two main categories: traditional opera motifs and auspicious symbols. These patterns are deeply reflective of both the era in which they were created and the local folk customs [41].

The main graphic serves as the central element of the pattern, typically depicting figures, animals, flowers, birds, insects, fish, lanterns, and other subjects. The auxiliary graphics complement the main graphic, enriching the overall composition and enhancing the visual atmosphere. These supplementary elements often include animals such as squirrels, foxes, elephants, and bats, as well as plants such as flowers, bamboo leaves, and lanterns. Geometric motifs are also commonly used as border decorations, contributing to the framing of the pattern. The divider line functions as a separator between the main graphic and the auxiliary elements. It typically forms enclosed geometric shapes such as circles, hexagons, octagons, or petal-like designs. Common geometric patterns in Blue Clamp-Resist Dyeing include the linked pearl pattern, cloud thunder pattern, and meander pattern, as shown in

Table 1. Deconstruction and analysis of Blue Clamp-Resist Dyeing patterns.

Dimensions	Main Graphic	Auxiliary Graphic	Divider Line
Pattern theme	Figures, animals, flowers, birds, insects, fish, lanterns, and other motifs.	Squirrels, foxes, elephants, sheep, bats, flowers, bamboo leaves, lanterns, and other animals, plants, and everyday objects.	Circular, hexagonal, octagonal, petal-shaped patterns; bead patterns, cloud and thunder patterns, Greek key patterns.
Pattern type
Shape	Concise and vivid, with a succinct summary; diverse and rich in content.	Abstract and concise, with rich content.	The shapes are diverse, divided into simple and complex forms. Most are composed of lines, with some patterns used as embellishments.
Composition	Mostly composed with a symmetrical arrangement on both sides.	Open composition, distributed in the four corners of the main graphic, mostly in a left-right symmetrical arrangement.	Mostly closed composition, with a left-right symmetrical arrangement.

.

4.3. Model Training

To ensure the authenticity and reliability of the model training, this study employed a comprehensive process of data screening, decomposition, preprocessing, and prompt word generation. Incomplete or unclear data were excluded, resulting in a final dataset of 132 high-quality images for training.

To maintain model accuracy, the images in the database were standardized before training. Given the large volume of data, the Birme 2.0 (Bulk Image Resizing Made) tool was used to batch-process the selected Blue Clamp-Resist Dyeing images. The images were resized to 512 pixels by 512 pixels, and the file format was standardized to PNG to ensure uniformity and compatibility with the training model. After preprocessing, the dataset required for model training was created. The images were then tagged using the WD1.4 Tagger tool, with manual optimization applied to ensure precision.

In the prompt word writing phase, this study developed a framework for descriptive prompts based on the structure, color schemes, and symbolism of Blue Clamp-Resist Dyeing patterns, ensuring accurate and detailed descriptions. To further enhance the effectiveness of the training model, negative prompts (e.g., “low quality” or “damaged”) were also included in the framework to prevent undesirable outputs.

The experimental environment for model training and image generation is shown in Figure 7.

After preparing the Blue Clamp-Resist Dyeing database and inputting it into the training set, the foundational large model for LoRA was selected. Key training parameters, including Repeat, Epoch, Batch Size, Learning Rate, and Optimizer settings [42], were adjusted to initiate the training process. The result was a trained LoRA model.

Upon completion of the training, the LoRA model was imported into Stable Diffusion for evaluation. The model’s effectiveness was analyzed using the X/Y/Z table. The final model was assessed based on its ability to replicate the desired Blue Clamp-Resist Dyeing style. In the evaluation, the X-axis represented the number of iterations (NUM), while the Y-axis indicated the weight (STRENGTH) of the model. A higher Y-axis value indicated a closer match to the original Blue Clamp-Resist Dyeing pattern. When the Y-axis weight was between 0.0 and 0.4, the model displayed little resemblance to the Blue Clamp-Resist Dyeing style. However, as the weight increased to between 0.8 and 1.0, the model began to incorporate distinctive elements of Blue Clamp-Resist Dyeing. At iteration two (X-axis), the model exhibited a hybrid style, blending the flat style of the large model with elements of Blue Clamp-Resist Dyeing. Between iterations four and eight (X-axis), the model showed a strong resemblance to Blue Clamp-Resist Dyeing, as shown in Figure 8.

4.4. Controllability Test

Based on the experimental results, this study employed the prompt words “two cute rabbits in the middle, floral and plant patterns in the four corners, butterflies, with floral and plant patterns on the four diagonals, symmetric pattern, flat, monochrome”, along with reverse prompt words such as “low quality, damaged, unclear”. The generated results are presented in Figure 9.

The test results demonstrate that the model effectively reproduced the style characteristics of Blue Clamp-Resist Dyeing patterns, aligning closely with the specified prompt. The model showed sensitivity to key design elements such as shape, color, pattern, and overall composition, achieving a harmonious balance between aesthetic features and the creative input derived from the prompts. These results indicate that the model can faithfully replicate the visual style of traditional Blue Clamp-Resist Dyeing while also generating innovative variations based on the established style.

4.5. Effectiveness Evaluation

To further validate the effectiveness of the LoRA model in generating Blue Clamp-Resist Dyeing patterns, we employed the Delphi method. This approach utilizes multiple rounds of expert feedback to gradually reach a consensus, ensuring the scientific and objective nature of the evaluation. For this study, ten experts were invited, including inheritors of Blue Clamp-Resist Dyeing craftsmanship, scholars specializing in intangible cultural heritage protection, technical experts in computer vision and artificial intelligence, and professionals in art design. All experts possessed extensive domain experience and a deep understanding of the cultural significance and technical characteristics of Blue Clamp-Resist Dyeing.

The evaluation process was conducted in two rounds, with expert ratings and opinions gathered through anonymous feedback in each round. In the first round, the experts conducted an initial assessment of the Blue Clamp-Resist Dyeing patterns generated by the LoRA model based on five core criteria: artistic expressiveness, cultural consistency, technical implementation, innovation, and practical value. The results showed that the LoRA model performed well in artistic expressiveness and technical implementation, scoring 7.8 and 8.2, respectively. However, it received lower scores for cultural consistency and innovation, with scores of 6.5 and 6.0. The consistency index (Kendall’s W) was 0.45, indicating significant divergence among expert opinions. The main recommendation was to include more traditional Blue Clamp-Resist Dyeing samples and increase the diversity of the generated works.

In the second round, based on the feedback from the first round, additional traditional Blue Clamp-Resist Dyeing works and new generated patterns were introduced for expert evaluation. By increasing the quantity and diversity of the samples, the comparison basis for the experts was more comprehensive. The results showed a greater consensus, with the consistency index (Kendall’s W) improving to 0.75. Artistic expressiveness and technical implementation scores rose to 8.5 and 8.6, while cultural consistency and innovation improved to 7.5 and 7.2, respectively. Practical value also reached 7.5. The overall weighted score for the LoRA model’s generated content was 7.9, indicating its high potential for both reproducing traditional Blue Clamp-Resist Dyeing craftsmanship and promoting innovation.

4.6. Development of Interactive Experience Projects

Supported by the Wenzhou Blue Clamp-Resist Dyeing Museum, this research developed an interactive experience application for Blue Clamp-Resist Dyeing, as shown in Figure 10. By encouraging public participation in the creation of Blue Clamp-Resist Dyeing patterns, the program aims to increase public awareness and promote the broader dissemination of this cultural tradition. The user interactive experience process is shown in Figure 11.

In this program, users can create patterns based on keywords and personal preferences, and share, save, or output their creations (Figure 12). This interactive process not only fosters user engagement but also deepens interest in exploring Blue Clamp-Resist Dyeing culture. Through this participatory approach, the research encourages the ongoing dissemination and innovation of intangible cultural heritage, providing new pathways for its preservation and development [43].

Upon participating in the Blue Clamp-Resist Dyeing interactive program, users are first introduced to the program’s interface, which presents a basic overview of its functions as well as core information about Blue Clamp-Resist Dyeing culture. Through text, images, and videos, users gain an initial understanding of the history and cultural significance of Blue Clamp-Resist Dyeing, sparking their interest in further engagement.

In the interactive interface, users can access the “Generate” button to reach the page connected to the trained LoRA model. On this page, users are prompted to input keywords into the system’s framework to generate personalized Blue Clamp-Resist Dyeing patterns. The keyword framework ensures that the generated patterns maintain both cultural relevance and aesthetic quality. This setup allows the system to balance user creativity with control over the quality and integrity of the patterns, preserving the artistic style and traditional characteristics of Blue Clamp-Resist Dyeing. Users can also adjust the keywords according to their preferences, creating patterns that meet their individual aesthetic needs.

Subsequently, users can share, save, or export their creations, completing a cycle that reinforces their sense of accomplishment and involvement. This process strengthens their sense of belonging to and identification with Blue Clamp-Resist Dyeing culture. Additionally, the program offers an option to print the created works, allowing users to transfer digital Blue Clamp-Resist Dyeing patterns to physical objects (such as clothing, stationery, etc.). This feature further extends the user experience, providing tangible forms for virtual creations and enhancing users’ engagement and value recognition of Blue Clamp-Resist Dyeing culture.

At the end of the process, the program encourages users to provide feedback on their interactive experience. They can suggest improvements or share their insights on Blue Clamp-Resist Dyeing culture. This feedback is collected through data analysis to help developers optimize future versions of the interactive design. The program also facilitates users’ interaction with other creators through community platforms, promoting further exchange and collaboration, which helps increase the widespread participation and dissemination of Blue Clamp-Resist Dyeing culture.

5. Discussion

This study systematically applies generative artificial intelligence, based on the LoRA model, to the digital preservation and interactive experience of Blue Clamp-Resist Dyeing patterns. By integrating generative AI with user interaction design, the research developed an interactive platform that allows users to input keywords to generate personalized Blue Clamp-Resist Dyeing patterns. This participatory tool not only enhances user engagement in the dissemination of Blue Clamp-Resist Dyeing culture but also promotes a new mode of interaction between intangible cultural heritage and the public, offering a valuable approach for cultural transmission and education [44]. The following answers to the research questions are provided.

This study investigates the challenges faced by Blue Clamp-Resist Dyeing in its transmission and development, establishing a digital database to preserve its patterns. The research identifies key obstacles in its inheritance and innovation, specifically in pattern design. Through a systematic classification of Blue Clamp-Resist Dyeing patterns, including their elements, meanings, and structures, the study constructs a comprehensive database. This database serves as a foundational resource, ensuring logical consistency and integrity in the preservation of these patterns. The analysis not only reveals the aesthetic and cultural significance of Blue Clamp-Resist Dyeing but also lays the groundwork for its future preservation and innovation.

With the rapid advancements in AI technology, this study explores the role of generative artificial intelligence technology in the innovation of Blue Clamp-Resist Dyeing patterns. Using the Low-Rank Adaptation (LoRA) method, the study develops a Blue Clamp-Resist Dyeing style model. This approach enables the intelligent design and generation of patterns, preserving cultural relevance while fostering innovation. The research establishes a prompt framework that ensures a balance between traditional cultural elements and innovative creativity in AI-generated patterns. This demonstrates the potential of generative artificial intelligence technology to revitalize Blue Clamp-Resist Dyeing, offering new avenues for its transmission and development, while validating the feasibility of generative artificial intelligence technology in innovating traditional cultural forms.

Building on the development of an efficient pattern innovation tool, this study emphasizes the importance of a participatory and collaborative mechanism in promoting the sustainable development of Blue Clamp-Resist Dyeing. By integrating generative artificial intelligence technology into museum interactive design, this research introduces new vitality into the cultural transmission and innovation of Blue Clamp-Resist Dyeing. The experimental results applied to the museum’s interactive program expand the scope of pattern creation, overcoming the limitations of traditional single-creator models. The program allows users to generate patterns based on personal preferences, fostering deeper cultural engagement. This interactive process not only enhances users’ sense of cultural belonging but also creates a full cultural experience loop, from creation to sharing and application.

Through the interactive program, users can save, share, and apply their generated patterns to various media, enriching their engagement with the culture. This immersive experience enhances cultural interaction, which in turn supports the sustainable development and innovation of Blue Clamp-Resist Dyeing. The study highlights the significant application potential of generative artificial intelligence technology in the transmission of traditional cultural heritage, demonstrating its value in intangible cultural heritage museum and exhibition contexts.

6. Conclusions

This study established a multidimensional digital database of Blue Clamp-Resist Dyeing patterns, providing data support for academic researchers and opening up a creative resource repository for designers. By developing a Blue Clamp-Resist Dyeing style LoRA and applying it to an intangible cultural heritage museum setting, this research demonstrated the efficiency and practicality of this technology in the innovative application of Blue Clamp-Resist Dyeing patterns, expanding the scope of potential applications. Through an immersive experience in creating Blue Clamp-Resist Dyeing patterns, users have disrupted the traditional models of dissemination and development of Blue Clamp-Resist Dyeing. This participatory, collaborative innovation mechanism has paved new pathways for the inheritance and development of Blue Clamp-Resist Dyeing patterns, injecting new momentum into the sustainable development of this intangible cultural heritage.

The results of this study significantly enrich the diversity and multiplicity of Blue Clamp-Resist Dyeing patterns, offering greater possibilities for the sustainable development of this intangible cultural heritage. By combining traditional cultural elements with modern technology, this research provides new perspectives and methods for the protection and innovation of intangible cultural heritage, promoting the organic integration of traditional craftsmanship and modern technology, and advancing further development in related fields.

However, this study still faces some limitations. In particular, the generative AI technology’s ability to reproduce the intricate and delicate patterns of Blue Clamp-Resist Dyeing needs improvement to ensure that innovative patterns not only maintain the traditional essence but also enhance their precision and creativity. Therefore, future research will focus on optimizing the AI model to improve its ability to handle complex patterns and deepen the exploration of innovative pattern strategies. The goal is to continuously inspire new design ideas while preserving their traditional essence, contributing more technological wisdom and artistic creativity to the protection and development of Blue Clamp-Resist Dyeing and, by extension, to other intangible cultural heritages.