In the Qing Dynasty, Guangzhou held significant prominence as a crucial point of origin for the South China Sea route of the Maritime Silk Road.The presence of paintings in the Guangdong region had a significant impact on the production of needlework products. Furthermore, it distinguishes Guangzhou embroidery as a distinct and noteworthy form of embroidery. The National Bureau of Statistics of China has recently published data indicating that the total exports of embroidered lace fabrics in the first quarter of 2023 amounted to 1,318,375 yuan, reflecting a cumulative growth of 6.6% compared to the corresponding period in the previous year^[1]. This suggests that the field of embroidery continues to possess untapped potential for further advancement. Consequently, a key concern lies in the effective transmission of Guangzhou embroidery techniques and their integration into contemporary embroidery applications, thereby facilitating the development of modern embroidery.

At present, Guangzhou Embroidery Craft Factory Co., Ltd. employs a limited number of embroidery craftsmen, numbering fewer than 10. Despite the efforts made by the Guangzhou government to foster the growth of Guangzhou embroidery, several challenges hinder the widespread adoption and preservation of this craft. These challenges include a limited number of skilled practitioners, the extensive training period required for embroidery craftsmen (typically exceeding five years), and the high cost associated with creating intricate designs. Consequently, Guangzhou embroidery works tend to be expensive, posing obstacles to their wider dissemination and utilization. In this study, we provide a proposed AI drawing training model aimed at facilitating Guangzhou embroidered sketch design. Additionally, we introduce Midjourney, an AI drawing system designed to enhance the convenience of Guangzhou embroidery sketch design.The utilization of AI drawing technology offers individuals lacking proficiency in drawing and design the capacity to produce satisfactory artwork by utilizing textual and visual resources. This aligns with contemporary technological advancements and transcends conventional human thought processes. Furthermore, it expands design and creative possibilities while partially addressing the challenges associated with the widespread preservation and application of Guangzhou embroidery works. The involvement of Chat-GPT, social groups, and practitioners is crucial in this process. By utilizing human-computer design, the study seeks to reduce the complexity associated with creating Guangzhouese embroidery sketch designs, thereby facilitating the development of various products such as bookmarks, silk scarves, cosmetic kits, charms, and DIY material kits.

The study of Guangzhou embroidery in China has centered on elucidating regional specificity and artistry. The Lingnan School, an important school of painting that emerged in modern China in the Guangdong region, was also an important source of paintings for Guangzhou embroidery, and was profoundly influenced by Ju Lian and the paintings are frequently based on flowers, birds, and fish, which are Hao analyzes the decorative techniques of more than 140 shawls and concludes that "fullness as beauty" is the artistic style of Guangzhou embroidery^[2]. Chen compared exported Guangzhou embroidery fabrics according to three categories: ornamental, custom-made, and bulk, in terms of spatial sense, patterns, and other artistic characteristics, reflecting the convergence of Chinese and Western artistic styles along the Maritime Silk Road and the diverse demands of overseas consumer markets during the period^[3].Along with the changing times and aesthetics, however, the formerly regionally distinct drafts gradually homogenized and lacked innovation, and Guangzhou embroidery inevitably confronted the same problem and lacked relevant studies. In terms of Innovativeness,, researchers have focused on development strategies. Lin and Wang evaluated the experience value of 20 participants in the Guangzhou embroidery experience program and proposed the first Guangzhou embroidery cultural consumption experience strategy based on the methodological purpose chain theory^[4]. The local embroidery industry in Suzhou responded actively to the changing market environment by increasing product segmentation and customization as well as intra/inter-industry cooperation, which helped the industry attract a large number of local talents and meet market demand, while preserving space for high-end artistic products, serving as a case study for the overall development and innovation of the Guangzhou embroidery industry^[5]. Guangzhou embroidery has the artistic characteristic of saturated color, but the traditional method of color perception and recognition of Guangzhou embroidery images has the problem of poor restoration ability. Therefore, a method of color perception and recognition of Guangzhouese embroidery images based on discrete mathematical models was proposed to facilitate the construction of a color database of Guangzhou embroidery and initiate the use of computers in the study of Guangzhou embroidery^[6]. Under the trend of smart wearable textile research, embroidery as an essential textile craft can be combined with technology to extend more possibilities^[7], and technology is one of the most important ways for Guang embroidery to make a breakthrough in innovation.

AI is presently used less frequently in the actual production of embroidery, but it does provide new opportunities for preserving and revitalizing traditional arts and crafts. Zhang employs the GANs model to analyze the New Year prints and the K-means algorithm to extract the color and colorize the prints from the Pop Art dataset in order to generate New Year prints with a Pop Art aesthetic^[8]. The aforementioned application of AI in traditional crafts provides a reference for the combination of Guangzhou embroidery and AI, but there is currently no survey on the popularity of traditional crafts created by combining AI to evaluate the appeal of traditional crafts combined with AI.

Traditional processes for formulating strategies are analytical, linear, problem-centric, and retrospective. They intend to capitalize on what is known by employing analytical and quantitative methodologies. Design thinking, on the other hand, prioritizes real people and emphasizes the significance of exploring the unknown^[9] and aligning individual objectives around a common outcome through team interaction^[10]. In 2005, the British Design Council introduced the Double Diamond model, which combines divergent and convergent thinking with step-by-step explanations and can be used as a more accessible simplification of the d.School's five phases of design thinking. The process paradigm for this study is based on the design thinking double diamond approach and its derived design thinking for strategy (DTS) design thinking model (Figure 1). In accordance with the two current development goals of Guangzhou embroidery, which are the need to differentiate it from other embroideries and its sustainable application in a modern society to attract younger groups, and based on its human-centered requirements, the opinions of various groups can be incorporated into the study.

In order to implement the matrix questionnaire design, the questionnaire contained 18 questions (Figure 2), and the four dimensions of regional specificity, innovativeness, artistry, and popularity were sorted out based on fieldwork and literature review, while the fifth dimension of understanding about Guangzhou embroidery was summarized based on the preliminary research. The purpose of the questionnaire was to determine if the AI-generated embroidery sketches were aesthetically pleasing, if they were recognizable and distinctive, and if respondents were willing to purchase related products, as well as to calculate the Customer Satisfaction Index (CSI) scores to quantitatively evaluate the sketches. The Customer Satisfaction Index is a comprehensive evaluation index of the level of satisfaction of representative customers of various types and levels, used to forecast industry trends influenced by changes in customer satisfaction^[11].

The research population comprised of 1) non-embroidery social groups, 2) students who had attended embroidery training classes, and 3) embroidery industry workers in order to obtain accurate and detailed data and content from a variety of perspectives.

The 124 Ju Lian paintings were submitted to Chat-GPT for prompts extraction around the five aspects of Theme, color, Style, Composition, and Atmosphere, but 14 failed to be identified due to insufficient resolution of the paintings.The descriptions of Chat-GPT are repetitive, and excluding the excessively simplistic and inaccurate descriptions, Table 1 summarizes the three dimensions other than Theme and color.

Due to the fact that Chat-GPT is insufficient for summarizing the two dimensions of Theme and color, it is currently impossible to determine the precise color parameters of specific species and targets; therefore, additional summaries based on field research are required.Table 2 classifies the major themes of Guangzhou embroidery into five categories. Due to translation and regional species, Paradisaea apoda and golden Arowana have been discovered, and AI recognition is currently insufficiently precise.Through field research in a Guangzhou embroidery craft factory, this study utilized a (Deta) digital colorimeter to extract color from a selection of Guangzhou embroidery works and silk threads (Figure 3) in an attempt to compile a color library for Guangzhou embroidery. Red, green, blue, and yellow were categorized as the four primary color families in order to select appropriate colors for implementation in this study and to provide a color reference for subsequent studies, including both RGB and CMYK data.

To ensure that AI can precisely identify the subject matter of large embroidery paintings, it is necessary to submit the corresponding theme in advance. Among them, the popular subject matter of lotus and butterfly is easier to recognize, and the combination does not require specific specific birds in Guangzhou embroidery as birds typically do, allowing AI's specialty to be performed more freely in the experiment for summarization. This study prioritizes the coupling of lotus and butterfly for practice in order to provide an example of how to apply the corresponding steps to other subjects.

In this study, the AI painting generation processes are primarily divided into four sequential steps: (1) Determine Theme, Style, and Composition, and select corresponding prompts from Tables 1 and 2. (2) Determine Atmosphere, and select pertinent prompts from Table 1. (3) Specify the Guangzhou embroidery color used. Select hue from Figure 3 (4) Painting cycle Generate four times and select the best results. Following is a detailed explanation of how artificial intelligence generates paintings.

Step 1: Determine the theme and composition style (Figure 4 left). Guangzhou embroidery is distinguished from other forms of embroidery by its use of themes with Lingnan characteristics, and the ability to accurately identify and generate target themes is particularly important in production practice. Second, the drawing's fundamental composition and style can have a significant impact on the overall visual effect and the embroidery work performed by the embroiderer. In this phase, the drawing is initially confirmed and evaluated by observing and comparing the results generated before and after adding the reference drawing, establishing the subject's composition, and establishing the basic colors.

When no theme color is specified and reference paintings are included (Figure 4, left), AI is able to accurately depict the subject matter based on the keywords and meet the partial overlay requirement. However, the overall tone is gloomy, the atmosphere is gloomy, and the background is too apparent. Although it can initially simulate the traces of ink painting, the ink pattern is not natural and realistic, which will not only result in the final presentation of the fabric not meeting the requirements of Guangzhou embroidery, but will also cause the embroiderer's quality of the fabric to deteriorate due to inaccuracy in judging the pattern of the painting's details. Priority is given to the simple color description of the painting's subject, the enhancement of the overall picture's luminosity, and the addition of Ju Lian painting to control the degree of haloing (Figure 4, right).

Step 2: Determine the atmosphere of the painting. It was discovered in Step 1 that the atmosphere can influence the overall feeling of a painting, and the Guangzhou embroidery focuses primarily on positive, well-wishes and fresh, tranquil images. The ambience adjectives provided by Chat-GPT in Table 1 were applied and compared (Figure 5), with the left side using words relatively favouring a serene atmosphere and the right side using words relatively favouring a positive atmosphere, both of which could fulfil the fundamental color use requirements. The results indicate that the error rate is relatively higher in paintings with a positive atmosphere, that the number of butterflies tends to exceed the minimum required to achieve the desired atmosphere, and that the form of the lotus exhibits a greater deviation, indicating that positive atmosphere requires more words for qualification and a greater degree of expansion. In this combination of lotus and butterfly, the ambiance is more suited to the painting's style and requirements, and the rate of accuracy is higher. The next stage is to complete the painting practise based on the tranquil environment.

Step 3: Determine the specific colors to be used for the Guangzhou embroidery. The silk thread color library is constructed according to the requirements of the woven products for each embroidery type. In this study, the corresponding RGB parameters were provided to AI as part of the keywords in order to make the overall artistic effect more appropriate for embroidery (Figure 6). It was discovered that the AI was capable of judging the color range on its own and would not override the colour parameter for the target as a whole when specific color parameters were inputted, affecting the visual effect as a whole. Consequently, additional settings can be made to integrate additional Guangzhou embroidery colours and bring the painting closer to the actual production specifications.

Through the three stages of determining the theme's style and composition, the painting's atmosphere, and the specific Guangzhou embroidery colours used, it has been possible to generate target sketches from five aspects: theme, composition, colour, style, and atmosphere. To demonstrate the validity of the keywords and the generated sketches, since Midjourney was able to provide 4 sketches each time, the final keywords were resubmitted 4 times, and the 1 sketch that best matched each time was chosen to produce 4 sketches as a result (Figure 7).

In order to ensure the validity of the model, based on the above categorization, this study temporarily selects 1-2 themes with obvious Guangzhou embroidery characteristics from each of the remaining categories for combining practise, again following the above three steps, containing key words in five dimensions: theme, style, composition, atmosphere, and color, and attempting to apply other Guangzhou embroidery colors to the target objects (Figure 8).

For the output AI drawings for questionnaire assessment, Bartlett's sphericity test was significant (χ2 (120) = 4162.097, p <.001); Kaiser-Meyer-Olkin value of 0.847 and sample size of 239 (130 males and 109 females), in line with Cronbach' s α coefficient value of 0.908, indicating that the reliability validity of the questionnaire met the criteria^[12].

A preliminary factor analysis was conducted in order to extract the appropriate number of factors and rank them according to the characteristic roots and post-rotation variance explanation rates. According to the table of total variance explained, the top 5 factor variables explained the characteristic root greater than 1. The 5 factors in the post-rotation variance explained a total of 90.2% of the total variance, factor 1 (Regional specificity) contributed 22.1%, factor 2 (Understanding) contributed 21.9%, factor 3 (Innovativeness) contributed 17.4%, factor 4 (Artistry) contributed 17.2%, and factor 5 (Popularity) contributed 11.6%, so it is recommended to retain the first five factors (see Table 3). Principal component analysis with the retained 5 factors showed that all questions met the common criteria (> 0.4) in terms of factor loadings and did not exhibit negative effects on each other or strong correlations (Table 4), so there was no need for question exclusion. Since the factor loading coefficient of PO-16 is 0.911 in Regional specificity, which is much larger than the coefficient in Popularity, the popularity of the drawing is deeply influenced by the region, and the question corresponding to factor 5 is more inclined to show the consumer willingness of AI painting after the introduction of wide embroidery, so the original fifth factor Popularity (PO) is changed to Consumer willingness(CW), Regional specificity(RS) was changed to Regional feature preference(RF), PO-16 was incorporated into Regional feature preference, and the corresponding Cronbach's alpha for the five dimensions was calculated. Cronbach's α to demonstrate the internal consistency. The alpha for the first factor (Regional feature preference ) = 0.955; the alpha for the second factor (Understanding) = 0.952; the alpha for the third factor (Innovativeness) = 0.948; the alpha for the fourth factor (Artistry) = 0.954; the alpha for the fifth factor ( Consumer willingness) with α= 0.927, these α values are greater than 0.8 and can be considered as good or higher^[13].

Understanding (items 1-4), Innovativeness (items 5-7), Artistry (items 8-10), Regional Feature Preference (items 11-14), and Consumer Willingness (items 15-16) are the outcomes of the final project.Regional feature preference refers to the embodiment and adoration of the trees, flowers, birds, and painting styles of Lingnan, China; Understanding refers to the degree of knowledge about Guangzhou embroidery; Innovativeness refers to the originality of incorporating artificial intelligence into the production and research of Guangzhou embroidery; and Artistry refers to the conformity to contemporary aesthetic and artistic characteristics.

For a comprehensive evaluation of the five dimensions, the Customer Satisfaction Index (CSI) uses the following standard method to compute the scores for each of the five dimensions.

Where η denotes the satisfaction of the research users, is the hidden variable, E (η), Max (η), Min (η) denote the expected, maximum and minimum values of overall customer satisfaction, respectively. This paper is a 7-level scale question, Min (η) = 0, Max (η) = 6. The overall satisfaction level is 61.05, which is an acceptable level (> 61), indicating that the combination of AI drawing and wide embroidery in this study is generally acceptable to young consumers and the study results are feasible. 5 dimensions of CSI scores are as follows:

Understanding is 58.3, Innovativeness is 63.1, Artistry is 62.1, Regional feature preference is 63.9, Consumer willingness is 57.6, Understanding, and Consumer willingness was low (≤61), while Innovativeness, Artistry, and Regional feature preference were acceptable (> 61). To ensure that AI painting has practical applications, two single options were introduced to the target consumer groups in Table 8 to clarify the direction of future research and application scenarios. In all aspects, the popularity of AI painting was relatively average, and there was no apparent frequency gap, indicating that AI painting in Guangzhou embroidery is more important for achieving an overall comprehensive effect than for determining individual preferences. In terms of product application preferences, the youth group selected products with greater practicality in life and relatively lower prices, while only 6.28 percent selected the current mainstream products of Guangzhou embroidery (traditional fine embroidery wall paintings), indicating that the needs of the youth group differ from those of the stable consumer group of Guangzhou embroidery at present, and overlap with the needs of the youth group from the previous study, who also desire products with greater practicality in life. They are also interested in purchasing a variety of useful embroidery products.

For the practical application of the drawings, nine industry practitioners, including six workers who had been engaged in Guangzhou embroidery for more than ten years, a sketch printer, and a Guangzhou embroidery salesperson (Table 5) evaluated 15 drawings from five categories. The overall opinion was divided into three parts.

(1)Considering that AI drawings can be transformed into Guangzhou embroidery works.

All six workers agreed that the 15 draughts could be used to complete the corresponding weaving, but because the draughts were still generated with slight structure deviations and a lack of clarity, the designers still needed to make adjustments to the AI drawings in the details, particularly in areas that could display the characteristic stitches of Guangzhou embroidery, such as flowers and feathers. In addition, after the overall attractiveness and clarity of the drawing have reached the corresponding standard, some details may need to be designed twice, as the AI's ink halo may still contain unfinished portions.

According to Liang Shu Ping (NO.6), an ICH inheritor, the AI sketches must be modified in accordance with the actual embroidery threads and stitching in order to adapt the structure of the sketches to the embroidery production method and enhance the sketches' aesthetic appeal.

Ms. Liu (NO.7), a printer, believes that AI draughts can solve the problem of colour difference of previous embroidery threads to some extent, but in the subsequent use, further modification of color parameters may still be required prior to the actual use, which is related to the printing machine in the factory, and frequently a painting must be adjusted and compared with silk threads before the final painting on silk or paper can be achieved.

(2)AI drawings of Guangzhou embroidery characteristics.

Guangzhou embroidery is characterised by embroidery techniques, which are more apparent in their application. The subjects in the sketches are ubiquitous in embroidery, but it is not sufficient to judge embroidery based solely on the sketches. Therefore, the designs and stitches must be combined to create the characteristics of embroidery.

Ms. Wu (NO.2) stated that the embroidery stitches should be matched to different parts of the painting, such as the embroidery of birds' feet and lychees, and only after the matching is completed can it be said that the work has the characteristics of embroidery from the drawing to the embroidery method and the whole.

(3) The application of AI drawing in broad embroidery.

Currently, AI paintings are better suited for comparatively low-end embroidery production in the Guangzhou embroidery industry, due to the industry's stable consumer base. This group of consumers prefers to collect embroidery works that can be augmented in value by both excellent non-genetic inheritors and renowned painters in order to increase the collection's value. AI paintings, on the other hand, can attract the younger generation and reduce costs to some degree, which is a realistic method to increase embroidery's consumer base.

Manager Zhang (NO.8) stated that in production practise, AI paintings are applicable to comparatively low-end products (partially decorated scarves and shawls, notebooks, postcards, fans, cloth stickers, DIY material kits) and have greater cultural dissemination significance. Due to the current shortage of Guangzhou embroidery workers, they are only able to satisfy the initial demand for hanging paintings; if AI painting is used for high-level hanging paintings, the workload will exceed the load, and it will be too labor-intensive.

The introduction of AI painting can attract many young people, which is beneficial for the promotion of Guangzhou embroidery's popularity and has a new way to attract the public on the new media platform, and it has the unique property of being able to generate corresponding drawings according to one's own needs, which is very interesting and corresponds well to the product of DIY material kit, as well as providing new inspiration for Guangzhou embroidery's products in printing.

Although AI has been gradually incorporated into the work and study of the general public, and the sustainable development of intangible cultural heritage requires identifying key points where new technologies can be incorporated, there is little research on the introduction of related technologies into traditional craft production, and embroidery still lacks a proper application of AI workflow and an appropriate research method to measure the feasibility of incorporating AI into embroidery. Consequently, the purpose of this study is to construct a process for the application of AI in Guangzhou embroidery, provide a basic reference process for practitioners, and initially verify the feasibility of AI painting in Guangzhou embroidery, provide dimensions to be noted for future real-world applications, and promote the positive role of AI painting in Guangzhou embroidery production in a number of ways.

This paper describes the process of establishing and validating AI as a new tool to assist Guangzhou embroidery in drawing design, proposes a working model combining the application of Chat-GPT and Midjourney for drawing design, and evaluates the regional feature from the perspective of the youth group by means of field research and questionnaire evaluation (16 questions and 5 subscales) preference, Understanding, Innovativeness, Artistry, and Consumer willingness, and Overall, based on sufficient preliminary research and practise, this workflow is sufficiently grounded for application and internally consistent for evaluation, thereby providing a sound fundamental research process for application and evaluation. Due to the limitations of this study, this study focuses on the working model of AI assisting Guangzhou embroidery to complete the drawing design; therefore, the CSI only reaches the acceptable standard and does not reach a high satisfaction score of 70 or more; however, the application of AI offers the potential for the sustainable development of Guangzhou embroidery and a distinguishing feature that can set it apart from other embroidery types. The existence of the five subscales and the CSI enables practitioners and researchers to assess the suitability of each generated drawing and make predictions about the future. We hope that additional researchers will build upon this foundation to further refine and expand how AI is applied to traditional crafts, to verify the feasibility of AI application to traditional crafts from additional perspectives, and to improve the generalizability of the scales by using more diverse samples.