AI-driven full marketing content design method for product packaging planning

Abstract

The present invention discloses an AI-driven full marketing content design method for product packaging planning, which relates to the field of marketing content design technology and includes the following steps: obtaining a product packaging design draft, extracting color, composition, font and semantic features based on a deep convolutional neural network, and fusing them to form a brand tonality mother vector of the brand style tonality; calling a multimodal analysis module to extract visual and language features consistent with the brand tonality mother vector dimension frame by frame or sentence by sentence from the multimodal marketing materials recently released by the brand, and generating a set of marketing material feature vectors. By constructing a brand tonality mother vector as a style anchor point, the present invention combines multimodal material analysis, a time-sensitive style evaluation mechanism and an adaptive style jump strategy, which not only achieves real-time monitoring of content style consistency, but also has the ability to dynamically adjust the style according to the risk of content fatigue, ensuring that the content generation system can intelligently perceive changes in the communication environment and make strategic responses.

Description

Technical Field

[0001] The present invention relates to the field of marketing content design technology, and in particular to an AI-driven full marketing content design method for product packaging planning. Background Art

[0002] AI-driven full marketing content design for product packaging planning uses product packaging design as the starting point and core basis for generating marketing ideas. Leveraging generative AI and multimodal content modeling technology, it automates the entire content creation process, from packaging to various marketing materials (such as promotional copy, visual posters, e-commerce detail pages, short video scripts, etc.). This approach extracts the brand's visual elements, language tone, and design style embedded in the packaging, driving the AI ​​system to generate stylistically consistent and semantically relevant multimodal marketing content. Adaptive optimization is then performed based on user profiles and market feedback, resulting in an integrated marketing content design model characterized by unified brand expression, efficient and collaborative content creation, and intelligently evolving communication strategies. This approach effectively breaks down the barriers between packaging design and marketing communications, transforming packaging from a mere physical carrier into a "style anchor" and "creative hub" that drives the entire marketing content planning and production process.

[0003] The existing technology has the following deficiencies:

[0004] In the process of AI-driven marketing content design for product packaging planning, to achieve a unified brand tone, existing technologies generally adopt fixed style transfer and content generation strategies, using the packaging visual style and language expression tone as a template basis to guide the generation of subsequent multimodal marketing content. However, in the process of pursuing content consistency, such methods often lack dynamic adjustment mechanisms for the diversity of typesetting structure, flexibility of color matching, and semantic level of copywriting. As a result, the generated graphic and text content tends to be homogenized in expression, making it difficult to present emotional tension and creative changes. With the repeated release of such content on multiple channels, users are prone to aesthetic fatigue and perceptual resistance, which in turn weakens the dissemination appeal of the content and the efficiency of interactive conversion on social platforms. At the same time, this problem also limits the system's ability to quickly adapt to changes in the packaging style of new products, affecting the brand's content response speed and continuous innovation capabilities in a dynamic marketing environment, becoming a major technical bottleneck restricting the large-scale application of AI-driven content systems.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an AI-driven full-marketing content design method for product packaging planning. By constructing a brand tonality mother vector and combining multimodal analysis, temporal evaluation and style jump mechanism, it can achieve consistency monitoring and dynamic regulation of content style, improve the intelligence and adaptability of content generation, enhance the attractiveness and interactive efficiency of brand communication, and extend the content life cycle to solve the problems in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an AI-driven full marketing content design method for product packaging planning, comprising the following steps:

[0008] S101. Obtain product packaging design drafts, extract color, composition, font, and semantic features based on a deep convolutional neural network, and fuse them to form a brand tonality vector of the brand style and tonality;

[0009] S102. Calling a multimodal analysis module to extract visual and language features consistent with the brand tonality matrix dimension from the brand's recently released multimodal marketing materials frame by frame or sentence by sentence, and generating a marketing material feature vector set.

[0010] S103. Calculate the cosine similarity between each marketing material feature vector and the brand tonality mother vector to obtain a style consistency score, and construct a style consistency score sequence in chronological order;

[0011] S104. Based on a preset temporal decay weight function, weight the style consistency score sequence to construct a dynamic consistency weight matrix, where new materials correspond to high weights and old materials correspond to low weights;

[0012] S105: merging the dynamic consistency weight matrix through an adaptive aggregation algorithm to obtain an initial style uniformity strength value, and normalizing and outputting a style uniformity coefficient;

[0013] S106. When the style unity coefficient exceeds a preset risk threshold, a multi-dimensional style space mapping diagram is constructed based on the current brand tonality state, a jump path is calculated according to the style unity coefficient, and the jump amplitude is dynamically adjusted in combination with the tonality unity strength.

[0014] Preferably, the step of generating the brand tone mother vector includes:

[0015] Obtain product packaging design draft images and input them into a pre-trained deep convolutional neural network model for preprocessing operations to extract color and composition features;

[0016] The text detection and recognition model is used to identify the text area in the packaging image, and the font classification network is used to extract font features;

[0017] The recognized text is input into the semantic embedding module to extract the language tone and emotional semantic features;

[0018] The color, composition, font and semantic features are fused and encoded, and weighted through the attention mechanism to generate the brand tone mother vector.

[0019] Preferably, the step of calling the multimodal analysis module to extract the marketing material feature vector includes:

[0020] Collect and standardize the images, videos, and text marketing materials released by the brand within a preset timeframe;

[0021] Sample the video material by frame, extract key frame images, and extract visual features through deep convolutional neural network;

[0022] Divide text materials into sentences and extract semantic sentiment and language tonal features through the natural language processing module;

[0023] Integrate image and text features in material order to generate a set of marketing material feature vectors that are consistent with the dimensions of the brand tone mother vector.

[0024] Preferably, the step of constructing a style consistency score sequence includes:

[0025] Perform L2 normalization on each feature vector in the marketing material feature vector set and the brand tone mother vector;

[0026] Call the cosine similarity function to calculate the similarity score between each feature vector and the brand tone mother vector;

[0027] Sort the similarity scores by the original release time of the materials;

[0028] Generate a sequence of style consistency scores in chronological order.

[0029] Preferably, the steps of constructing a dynamic consistency weight matrix include:

[0030] Calculate the relative time offset based on the timestamp information of each item in the style consistency score sequence;

[0031] Call the time decay weight function model to calculate the time decay weight for each score;

[0032] Multiply each style consistency score by the corresponding weight value to generate a weighted score;

[0033] Map all weighted scores into a set of dynamic consistency weight matrices arranged in order of materials.

[0034] Preferably, the temporal decay weight function is an exponential decay function, and its expression is W(t) = e^(-λt), where t is the time offset and λ is the decay coefficient. The exponential decay function is used to ensure that new materials correspond to high weights and old materials correspond to low weights, thereby realizing the response priority of content style evaluation to time changes.

[0035] Preferably, the step of generating the style unity coefficient includes:

[0036] Receive and parse the dynamic consistency weight matrix to extract the temporal weighted feature scores of each style sub-dimension;

[0037] Perform weighted fusion or nonlinear merging on the eigenvalues ​​of each column to obtain the initial style uniformity strength value;

[0038] A normalization algorithm is used to map the initial intensity value to a style unity coefficient between 0 and 1;

[0039] Store style consistency coefficients in the content monitoring dashboard for real-time viewing and analysis.

[0040] Preferably, the steps of calculating the jump path using the style unity coefficient and dynamically adjusting the jump amplitude based on the tonality unity strength include:

[0041] When the style unity coefficient exceeds the preset risk threshold, a multi-dimensional style space mapping diagram is constructed based on the brand tone status;

[0042] Determine the current style coordinates in the style space according to the style unity coefficient and calculate the jump path;

[0043] Dynamically adjust the jump range based on the strength of brand tonality to avoid tonality fragmentation;

[0044] The style parameters of the content generation model are updated according to the jump coordinates to achieve differentiated style control and creative guidance of subsequent content.

[0045] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0046] By constructing a brand tonality matrix as a style anchor, combined with multimodal material analysis, a time-sensitive style assessment mechanism, and an adaptive style jump strategy, this method not only enables real-time monitoring of content style consistency but also dynamically adjusts style based on content fatigue risk, ensuring that the content generation system can intelligently perceive changes in the communication environment and respond strategically. This method significantly improves the flexibility and intelligence of content creation, enhances the brand's content appeal and user interaction efficiency in multi-channel communication, effectively extends the lifecycle and communication value of brand content, and has outstanding practicality and broad promotion prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0048] Figure 1 This is a flow chart of the method for the AI-driven full marketing content design method for product packaging planning of the present invention. DETAILED DESCRIPTION

[0049] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0050] The present invention provides Figure 1 The AI-driven full marketing content design method for product packaging planning shown in the figure includes the following steps:

[0051] S101. Obtain a product packaging design draft, extract color features, composition features, font features, and semantic features from the packaging design draft based on a deep convolutional neural network model, and fuse the multiple visual and language features extracted above to construct a brand tonality vector representing the brand style and tonality.

[0052] The brand tonality matrix generation process includes the following four sub-steps, which are used to extract representative brand visual and linguistic style features from the product packaging design draft to construct a high-dimensional, semantically consistent tonality representation vector, thereby providing basic support for subsequent marketing content generation and style consistency assessment:

[0053] An image file of the target product's packaging design is obtained and fed into a pre-trained deep convolutional neural network model for image preprocessing, including size normalization, color standardization, and background noise removal. The preprocessed packaging design image is then fed into a backbone feature extraction network comprised of multiple convolutional and pooling layers to automatically learn and extract low-level color distribution features (such as dominant hue, saturation, and brightness ratio) and mid-level compositional structural features (such as layout, spatial hierarchy, and edge structure). These features are then combined with multi-scale convolutional kernels to enhance spatial perception, ensuring the model can identify stylistically representative compositional forms and color patterns within the image.

[0054] To capture font features in packaging images, this paper introduces an auxiliary approach based on a joint text detection and recognition model. This approach locates, segments, and identifies text areas within packaging images. The identified text content is then fed into a pretrained font classification network to further extract font style features, such as type (serif / sans serif), weight, slant, and stroke structure. These features are then associated and encoded with the original image features. The font features captured through this approach not only reflect the visual communication language of the packaging design but also complement the unified representation of the brand style through font expression.

[0055] In order to extract semantic features, the present invention inputs the packaging text content identified in the second step into a semantic embedding module based on the Transformer structure through a joint language-visual multimodal alignment model to extract the language tonality and semantic emotional features expressed by it. The semantic embedding module supports the recognition of subjective style features such as "youthful vitality", "sense of technology", "warmth and friendliness", and realizes collaborative encoding from visual language to semantic style by fusing and mapping with image features extracted from the visual channel. The introduction of semantic features enables the extracted brand tonality representation to have not only visual style characteristics at the sensory level, but also language consistency description capabilities at the content level.

[0056] The color, composition, font, and semantic feature vectors extracted from these three channels are then jointly encoded in a high-dimensional manner using a feature-level fusion network. An attention mechanism is then introduced to dynamically assign weights to the features across each dimension, ultimately forming the final brand tone matrix. This matrix, a high-dimensional representation that integrates visual and linguistic style features, possesses excellent style abstraction capabilities and consistent expression.

[0057] This step extracts and fuses multi-dimensional deep features from the product packaging design to construct a high-dimensional vectorized representation that comprehensively characterizes the brand's visual and linguistic style and tonality, known as the brand tonality matrix. Product packaging is often one of the most representative visual symbols in brand communication, embodying the brand's core style, aesthetic tendencies, and communication context. By extracting visual features such as color, composition, and font from packaging images based on a deep convolutional neural network model, and combining them with the semantic tonality inherent in the text within the image, structured modeling of brand tonality can be achieved. The fusion of these visual and linguistic features not only enhances the model's ability to consistently express brand style but also provides a standardized and quantifiable "style anchor" for subsequent marketing content generation.

[0058] S102. Calling the multimodal analysis module to process the brand's recently released multimodal marketing materials, extracting visual and language feature vectors that are consistent with the brand's tonality matrix vector dimension frame by frame (for video materials) and sentence by sentence (for text materials), and generating a set of marketing material feature vectors in the order of the materials.

[0059] In this embodiment, the step of invoking the multimodal analysis module to extract the feature vector of brand marketing materials aims to extract unified semantic and stylistic features from various marketing materials (including images, videos, text, etc.) released by the brand across different media channels, in order to match and compare them with the brand tonality matrix vector. This step includes the following specific operational procedures:

[0060] The system collects data on all marketing materials released by a brand within a preset time window (e.g., the past 30 days). This includes video ad clips, social media image and text push notifications, e-commerce detail page images, product short videos, and various types of copy. For video materials, the system evenly samples each frame, extracting keyframe images at regular intervals. For text materials, the system performs sentence-by-sentence structured analysis, treating each sentence as an independent unit for analysis. After standardization (including size unification, frame rate normalization, and language cleaning), these materials are fed into the multimodal analysis module for feature extraction.

[0061] For keyframes in images and videos, a deep convolutional neural network (DCNN) with the same structure as the packaging design feature extraction model is used to extract visual features from each frame. This includes the image's primary color distribution, composition rules, background complexity, visual center positioning, and font layout characteristics. The dimensions of these visual features are strictly aligned with the visual dimensions of the brand tonality matrix, ensuring subsequent one-to-one style similarity calculations. For images containing text, OCR technology is used to identify the text and semantically co-encode it with the image features.

[0062] For text-based materials (including short copy, e-commerce headlines, and detailed description paragraphs), the system uses a natural language processing module to perform semantic embedding on each sentence, extracting features such as the sentence's semantic sentiment, linguistic tone, keyword distribution, and linguistic structural complexity. Built on a Transformer architecture, this natural language processing module aligns with the dimensionality of the linguistic feature vectors in the brand tonality matrix, ensuring that the multi-layered expression of text tonality is accurately embedded in a unified feature space. Semantic feature extraction not only incorporates thematic information at the content level but also incorporates subjective stylistic factors such as sentence structure and emotional tension, thereby achieving a comprehensive mapping of semantic tonality and visual style.

[0063] The feature vectors of each extracted material unit (frame image or sentence) are integrated in the order of material source to generate a complete set of marketing material feature vectors. Each feature vector maintains dimensional and structural consistency with the brand tone mother vector, enabling subsequent similarity comparison, style consistency scoring, and dynamic style control. This feature set not only preserves the visual and linguistic expressive characteristics of the content but also embeds the temporal dimension of the material. This serves as the fundamental input for style monitoring and content optimization throughout the content system, enabling cross-modal, cross-platform, and cross-temporal brand style consistency modeling and intervention control.

[0064] This step uses a multimodal parsing module to extract refined features and standardize representations of the brand's recently released marketing content. This creates a "marketing material feature vector set" that is strictly aligned with the brand's tonality matrix, providing the foundation for style consistency assessment and dynamic content regulation. In practice, brand marketing content often exists in multimodal formats, such as video, images, and short copy, each of which carries the externalization of visual style and linguistic tonality. This step first processes video footage frame by frame, extracting visual features (including color, composition, and font layout) from each keyframe, and then recognizes and encodes the text within the image. Second, it analyzes text sentence by sentence, extracting features such as semantic sentiment, sentence structure, and tonal tendencies within the language. This "cross-modal + unified dimensionality" approach enables the system to construct a structured, unified model for massive amounts of heterogeneous content, facilitating subsequent similarity calculations with the brand's tonality matrix. This step not only ensures the accuracy and consistency of content style analysis, but also provides a quantifiable and traceable data basis for subsequent style consistency scoring, style fatigue monitoring, creative style jumps, etc., thereby realizing style perception, intelligent decision-making and dynamic control of the entire marketing content process.

[0065] S103. Calculate the cosine similarity between each feature vector in the marketing material feature vector set and the brand tonality mother vector to obtain style consistency scores for multiple single materials, and construct a style consistency score sequence in chronological order of the materials;

[0066] In this embodiment, the steps of calculating the cosine similarity between each feature vector in the marketing material feature vector set and the brand tonality mother vector to obtain a style consistency score and constructing a score sequence in chronological order specifically include the following sub-steps, which aim to achieve quantitative evaluation of content style fit and time series modeling, so as to facilitate subsequent trend monitoring and dynamic style regulation.

[0067] For the constructed set of marketing material feature vectors, each feature vector unit is traversed in turn, and the cosine similarity calculation function is called to calculate the similarity between the feature vector and the pre-generated brand tone mother vector. The cosine similarity function calculates the cosine of the angle between two vectors in vector space and outputs a similarity value between 0 and 1, which is used to measure the degree of consistency between the current material's stylistic characteristics and the overall brand tone. The closer the value is to 1, the closer the material's visual and linguistic style matches the brand tone. This calculation method is directionally sensitive and can effectively capture structural differences between stylistic features, making it particularly suitable for brand style consistency analysis.

[0068] To ensure computational accuracy and dimensionality uniformity, the present invention performs L2 normalization on all eigenvectors before entering the similarity calculation to eliminate the effects of varying vector moduli. This normalization step not only improves computational stability but also provides a mathematical basis for fair comparisons between eigenvectors of different modalities. Projecting the normalized vectors onto the unit sphere results in more accurate similarity calculations, avoiding biases caused by content complexity or modality differences.

[0069] After calculating the similarity of each creative unit, the resulting style consistency scores are sorted by the original release time of the creatives and stored in a style consistency score sequence in this chronological order. This temporal sorting mechanism ensures that subsequent analysis of style evolution trends has a temporal reference value, helping to identify fluctuations in style consistency across different release cycles. Each score carries a timestamp, which is used to construct the subsequent dynamic weighting matrix and attenuation function model.

[0070] After traversing and calculating the entire collection of creatives, the system outputs a complete sequence of style consistency scores, which serves as the input for the brand tone monitoring system. This sequence allows operators to digitally track the degree of alignment between each piece of content and the brand tone. This can be further used to build style monitoring dashboards, trigger creative control mechanisms, or generate content fatigue risk alerts.

[0071] This step calculates the cosine similarity between each vector in the set of marketing creative feature vectors and the parent vector of the brand's tonality, quantitatively assessing the stylistic consistency between a single marketing creative and the overall brand's tonality. A chronologically ordered sequence of style consistency scores is constructed, providing a data foundation for subsequent style trend analysis, dynamic monitoring, and content strategy adjustments. In a brand's multi-platform, high-frequency marketing campaign, the stylistic consistency of its various creatives directly impacts user brand recognition and content acceptance. Traditional methods struggle to systematically measure this stylistic consistency. However, this step, by calculating the cosine of the angle between the feature vectors, accurately measures the degree of match between the creative and the brand's tonality. The resulting similarity score, ranging from 0 to 1, allows the system to capture stylistic divergence, change, or convergence in real time. Furthermore, by sorting creatives by release date and constructing a style consistency score sequence, the system continuously monitors the evolution of brand content style over time, providing a timely basis for the subsequent introduction of time-decrease weighting and jump triggering mechanisms. This step not only enhances the objectivity and operability of content consistency evaluation but also lays a solid foundation for efficient and flexible content style regulation.

[0072] S104: Based on a preset temporal decay weight function, weight each score value in the style consistency score sequence to construct a dynamic consistency weight matrix, where newer materials have higher weights and older materials have lower weights;

[0073] In this embodiment, the step of weighting each score in the style consistency score sequence based on a preset time-decay weight function to construct a dynamic consistency weight matrix aims to establish a time-sensitive style perception mechanism, enabling the system to more reasonably measure and respond to the evolution of brand content style in different time dimensions, thereby achieving dynamic and precise control of style monitoring and regulation. This step includes the following key operational processes:

[0074] Based on the style consistency score sequence constructed in the previous step, the system sorts the scores according to the timestamp information of the material and binds a corresponding time offset to each score. This time offset is a relative time difference calculated backwards from the current system time as the reference point, and can be expressed in hours, days, or weeks. This time offset information will be used as an input variable for the subsequent time series decay function to apply decay processing to each score value. By establishing a mapping relationship between each score item and its corresponding time difference, the system can achieve accurate time perception characteristics, providing a basis for dynamic weight allocation.

[0075] The preset temporal decay weight function model is called to calculate the weight of each style consistency score item. This weight function can take the form of an exponential decay function, a hyperbolic decay function, or an adjustable power decay function. The specific function expression can be set as W(t) = e^(-λt), where t is the relative time offset value and λ is the decay coefficient, which controls the speed of weight decrease. Newer materials have smaller t values, so the corresponding weight value is close to 1, indicating that their influence in style judgment is higher; while older materials have larger t values, so their weight value approaches 0, and their influence gradually decreases. The design of this function ensures that the content evaluation responds to the "current style state" first, giving the system the ability to perceive timeliness and quickly adapt.

[0076] According to the time decay weights calculated above, they are applied to each score item in the style consistency score sequence, and a new weighted score sequence is generated by weighting each item. Specifically, the system performs a vector-level multiplication operation, multiplies each score by its corresponding time decay weight, and outputs a set of weighted score values. Subsequently, the weighted score sequence is mapped to a multidimensional weight matrix. Each row in the matrix structure represents a material sample, and each column corresponds to a feature dimension or tonal factor. It can be extended to data alignment in multiple dimensions such as image style, language tone, and visual composition. In the matrix structure, the row vectors of newer materials have a higher overall weight contribution, while the influence of older materials is automatically weakened by the system due to weight decay.

[0077] The constructed dynamic consistency weight matrix serves as input for subsequent aggregation operations and style transition judgment. This matrix not only preserves the semantic information and structural distribution characteristics of the style consistency scores but also introduces a dynamic weight adjustment mechanism that changes over time, significantly improving the real-time nature and strategic flexibility of style judgment. Furthermore, the weight patterns in the matrix can be used as a basis for content lifecycle analysis, identifying active areas and risk zones of style fluctuation within a specific period, providing strategic reference for content creative planning.

[0078] This step introduces a preset time-decay weighting function to weight the style consistency score sequence for time sensitivity, thereby constructing a dynamic consistency weight matrix. This enables dynamic modeling and real-time monitoring of the style status of brand content. In actual marketing content management, materials released at different times have varying impacts on brand tone. Especially in the context of high-frequency, multi-platform content delivery, recent materials are more representative of user perception and brand image. By assigning higher weights to newer materials and lower weights to older materials, this step enables the system to prioritize content performance within the current time window when assessing style consistency, enhancing its timeliness. Furthermore, the constructed dynamic consistency weight matrix not only preserves the structural information of the original content style scores but also, through a time-weighted mechanism, reflects the trend direction of style change and areas of risk concentration. This provides fundamental data support with a temporal dimension for subsequent modules such as style aggregation, risk monitoring, and redirect triggering, effectively improving the accuracy and intelligence of content style management.

[0079] S105: merging multiple feature dimensions of the dynamic consistency weight matrix using an adaptive aggregation algorithm to obtain an initial style uniformity strength value, normalizing the initial style uniformity strength value to output a style uniformity coefficient with a value range between 0 and 1, and storing the style uniformity coefficient in a content monitoring dashboard for real-time viewing;

[0080] In this embodiment, the steps for merging the multiple feature dimensions of the dynamic consistency weight matrix using an adaptive aggregation algorithm to generate a style consistency strength value and normalize it into a style consistency coefficient specifically include the following operational process. This step aims to integrate the style consistency scoring results distributed across different modalities and dimensions into a quantifiable, monitorable, and traceable core indicator to comprehensively reflect the degree of consistency in the visual and linguistic style of the current brand content, providing a real-time reference for content operations and style control.

[0081] The system receives and parses the dynamic consistency weight matrix generated in the previous processing step. Each row in the matrix structure corresponds to a single material sample feature vector with temporal weight correction, and each column represents a specific style sub-dimension (such as color coordination, composition similarity, font style matching, semantic tone fit, etc.). Each value in the matrix has combined cosine similarity and time decay factors, which can truly reflect the difference in contribution of each dimension in different time segments. In order to unify the style evaluation standards, this implementation method introduces an adaptive aggregation algorithm, which has the ability to fuse multi-dimensional features and can dynamically adjust the fusion weights of features of different dimensions according to the distribution of input data.

[0082] A feature merging operation is performed on the dynamic consistency weight matrix. During this process, each column of features is first weighted averaged or nonlinearly weighted (e.g., using attention mechanism weighting, information entropy adjustment factors, or principal component weight extraction based on the covariance matrix) to obtain a representative score for each style sub-dimension. The scores of multiple dimensions are then combined and merged using an adaptive aggregation algorithm. The aggregation algorithm can use weighted averaging, fuzzy logic fusion, neural network regression, or a linear regression model based on multi-factor evaluation. The core feature of this algorithm is that it automatically adjusts the fusion strategy of each dimension based on the content distribution. For example, when the current semantic tonality differs significantly but the visual style is highly consistent, the system can appropriately increase the weight coefficient of the semantic dimension to more accurately reflect the "global unity" of the style.

[0083] After completing the aggregation of all dimensions, the system outputs an initial style uniformity strength value. This value is usually an unnormalized floating-point value, which indicates the degree of fit between the current content set as a whole and the brand tone. In order to improve the operability and monitoring convenience of cross-project comparisons, the present invention further designs a normalization mapping mechanism. This mechanism converts the initial strength value into a standardized indicator between 0 and 1 based on the maximum-minimum normalization method or Z-Score normalization strategy of sample data. The normalization process not only eliminates the impact of data scale differences on the final judgment, but also facilitates horizontal comparison with the uniformity data of other brands or time periods on the dashboard.

[0084] The normalized style consistency coefficient is written to a visual monitoring dashboard via the content interface and displayed in real time as a core indicator of brand content style consistency. The dashboard supports multiple visualization modes, including trend charts, heat maps, and interval distribution charts, making it easy for brands, content operators, and creative teams to track style fluctuations over time and identify risk signals of overly concentrated or dispersed styles. If the style consistency coefficient is consistently too high, the dashboard will issue a style fatigue risk warning, prompting the content system to trigger a style jump mechanism. If it is too low, it indicates a lack of style consistency, affecting brand recognition.

[0085] This step intelligently aggregates the style feature scores across multiple dimensions in the dynamic consistency weight matrix to calculate a quantitative metric, the style consistency coefficient, that comprehensively reflects the degree of stylistic consistency across a brand's current content. Since marketing content often encompasses multiple modalities and stylistic features, such as visuals, language, and typography, relying solely on a single dimension cannot accurately assess the overall consistency of a brand's tone. Therefore, this step incorporates an adaptive aggregation algorithm that dynamically assigns fusion weights based on the importance and diversity of different feature dimensions, enabling a comprehensive analysis of multidimensional style information. The initial style consistency strength value obtained after aggregation is normalized and mapped to a standardized coefficient between 0 and 1, ensuring excellent horizontal comparability and time-series visibility. Finally, this style consistency coefficient is logged in real time to a content monitoring dashboard, allowing brand operators to review and analyze the stylistic consistency of their current content at any time. This approach not only enhances the scientific and systematic nature of style management but also provides key decision-making support for subsequent style transitions and creative strategy optimization, significantly enhancing the intelligent control capabilities of brand content.

[0086] S106. When the style unity coefficient exceeds a preset risk threshold, a multi-dimensional style space mapping diagram is constructed based on the current brand tonality state, and a jump path of the style coordinates is calculated in the style space according to the style unity coefficient value. The jump range of the style coordinates is dynamically adjusted in combination with the current brand tonality unity strength, thereby achieving differentiated control and creative guidance of the subsequent content style.

[0087] In this embodiment, when the style consistency coefficient exceeds a preset risk threshold, the system constructs a multidimensional style space mapping diagram and calculates a style transition path to achieve differentiated control and creative guidance for subsequent content styles. Specifically, the following steps are included. This step aims to address the aesthetic fatigue and decreased communication efficiency caused by the long-term over-consistency of brand content styles. By dynamically transitioning within the style space, it rationally guides subsequent content to moderate changes, thereby enhancing the vitality and user appeal of brand communication.

[0088] When the system detects that the style consistency coefficient exceeds a preset risk threshold (for example, greater than 0.85), it indicates that the brand's recently released content exhibits excessive convergence in visuals, language, and typography. This triggers a style jump warning mechanism, constructing an interpretable and controllable multidimensional style space mapping based on the brand's historical tonality data, target audience profile, product attributes, and distribution channels. This style space uses various stylistic feature dimensions as axes, such as color tendency (warm / cold, neutral), visual tension (contrast, dynamic), composition (symmetrical / asymmetrical), language style (rational / emotional), and semantic emotion (positive / neutral / suggestive). This style space is constructed using an embedding model trained with historical marketing materials and style label data to achieve continuous expression of brand tonality in this multidimensional space.

[0089] Based on the current value range of the style unity coefficient, the style center coordinates of the brand's current content are determined in the style space. The direction and path of the style coordinate jump are calculated based on the extent of the risk threshold exceeded. The jump path calculation mechanism is based on the principles of maximizing differences and controlling tonal boundaries. That is, while ensuring that the brand's core tonality is not deviated from, the target area with a certain "stylistic distance" from the current style state is selected as the jump direction. For example, if the current style appears overly calm and conservative (such as low-saturation color schemes and overly rational language), the system will choose to jump to a more lively or emotional area in the style space, thereby achieving a localized shift in the tonal dimension and increasing the expressive tension.

[0090] To prevent tonal fragmentation caused by excessive style shifts, the system introduces the strength of brand tonal unity as a shift adjustment factor, dynamically adjusting the magnitude of style shifts based on this metric. Tonal unity strength can be determined through historical stability analysis and user cognitive feedback modeling, reflecting the brand's ability to maintain consistent tonal consistency. If the brand's tonal stability is high, the system allows for medium or large shifts within the style space to increase creative tension. Conversely, if the brand's tonal sensitivity is high, the system automatically limits the magnitude of the shift, making only minor adjustments within adjacent style areas to prevent brand tonal fragmentation or user cognitive misalignment. This dynamic control mechanism ensures that style shifts are not only directional and purposeful, but also controllable and flexible, ensuring the sustainable evolution of creative change within the framework of brand strategy.

[0091] Based on the calculated transition coordinates, the style parameters of subsequent content generation models are updated, including but not limited to content template selection, image generation style, text generation corpus bias, and color scheme changes. This automatically aligns subsequent marketing materials, including graphics, copywriting, and short video scripts, with the new style coordinates. Every style change during the transition process is recorded and visualized on the brand content style monitoring dashboard, enabling the operations team to evaluate the effectiveness of adjustments, user feedback, and changes in content click-through rates, achieving closed-loop data feedback and optimization.

[0092] When the system detects that the style consistency coefficient exceeds a preset risk threshold—indicating excessive consistency and convergence across visual, linguistic, and typographical aspects of brand content—the system triggers a timely style shift mechanism. By constructing a multidimensional style space map and dynamically adjusting coordinates, it guides subsequent content to achieve appropriate stylistic shifts and creative updates. Specifically, this step first constructs a visual and computable style space across multiple stylistic dimensions (such as color warmth and coldness, semantic mood, compositional style, and language style) based on the historical evolution of brand tonality, current content status, and user acceptance thresholds. Next, based on the value and trend of the style consistency coefficient, the system determines whether the brand content has fallen into style fatigue. If so, it calculates a reasonable style coordinate shift path. The system then incorporates the dynamic factor of brand tonality consistency strength to adjust the magnitude and direction of the shift, ensuring that stylistic differentiation is achieved while maintaining the core consistency of the brand tonality. Ultimately, this step drives the content generation system to moderately introduce new style factors, breaking the cycle of homogenized content and enhancing its freshness, emotional intensity, and communication appeal while preserving brand recognition and achieving a dynamic balance between "unity" and "diversity." This intelligent style transition mechanism effectively addresses the problem of traditional AI-generated content lacking in variety and maintaining user engagement after multiple rounds of delivery. It is a key technical means of enhancing brand communication and content lifecycle.

[0093] The above-mentioned AI-driven full-marketing content design method for product packaging planning can effectively solve key problems in existing technologies such as content style homogeneity, user aesthetic fatigue, and slow system response, and achieve brand content creation while maintaining a unified tone while introducing dynamic changes and creative diversity. This method constructs a brand tone mother vector as a style anchor, combines multimodal material analysis, a time-sensitive style evaluation mechanism, and an adaptive style jump strategy, not only achieving real-time monitoring of the content style consistency status, but also has the ability to dynamically adjust the style according to the risk of content fatigue, ensuring that the content generation system can intelligently perceive changes in the communication environment and respond strategically. Ultimately, this method improves the flexibility and intelligence level of content creation, enhances the brand's content appeal and user interaction efficiency in the multi-channel communication process, significantly extends the life cycle and communication value of brand content, and has strong practicality and promotion prospects.

[0094] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. An AI-driven full marketing content design method for product packaging planning, characterized by: The following steps are involved: S101. Obtain product packaging design drafts, extract color, composition, font, and semantic features based on a deep convolutional neural network, and fuse them to form a brand tonality vector of the brand style and tonality; S102. Calling a multimodal analysis module to extract visual and language features consistent with the brand tonality matrix dimension from the brand's recently released multimodal marketing materials frame by frame or sentence by sentence, and generating a marketing material feature vector set. S103. Calculate the cosine similarity between each marketing material feature vector and the brand tonality mother vector to obtain a style consistency score, and construct a style consistency score sequence in chronological order; S104. Based on a preset time-series decay weight function, weight the style consistency score sequence to construct a dynamic consistency weight matrix; S105: merging the dynamic consistency weight matrix through an adaptive aggregation algorithm to obtain an initial style uniformity strength value, and normalizing and outputting a style uniformity coefficient; S106. When the style unity coefficient exceeds a preset risk threshold, a multi-dimensional style space mapping diagram is constructed based on the current brand tonality state, a jump path is calculated according to the style unity coefficient, and the jump amplitude is dynamically adjusted in combination with the tonality unity strength.

2. The AI-driven full marketing content design method for product packaging planning according to claim 1 is characterized in that: The steps for generating the brand tone mother vector include: Obtain product packaging design draft images and input them into a pre-trained deep convolutional neural network model for preprocessing operations to extract color and composition features; The text detection and recognition model is used to identify the text area in the packaging image, and the font classification network is used to extract font features; The recognized text is input into the semantic embedding module to extract the language tone and emotional semantic features; The color, composition, font and semantic features are fused and encoded, and weighted through the attention mechanism to generate the brand tone mother vector.

3. The AI-driven full marketing content design method for product packaging planning according to claim 1 is characterized in that: The steps of calling the multimodal analysis module to extract the marketing material feature vector include: Collect and standardize the images, videos, and text marketing materials released by the brand within a preset timeframe; Sample the video material by frame, extract key frame images, and extract visual features through deep convolutional neural network; Divide text materials into sentences and extract semantic sentiment and language tonal features through the natural language processing module; Integrate image and text features in material order to generate a set of marketing material feature vectors that are consistent with the dimensions of the brand tone mother vector.

4. The AI-driven full marketing content design method for product packaging planning according to claim 1 is characterized in that: The steps for constructing a style consistency score sequence include: Perform L2 normalization on each feature vector in the marketing material feature vector set and the brand tone mother vector; Call the cosine similarity function to calculate the similarity score between each feature vector and the brand tone mother vector; Sort the similarity scores by the original release time of the materials; Generate a sequence of style consistency scores in chronological order.

5. The AI-driven full marketing content design method for product packaging planning according to claim 1 is characterized in that: The steps to construct a dynamic consistency weight matrix include: Calculate the relative time offset based on the timestamp information of each item in the style consistency score sequence; Call the time decay weight function model to calculate the time decay weight for each score; Multiply each style consistency score by the corresponding weight value to generate a weighted score; Map all weighted scores into a set of dynamic consistency weight matrices arranged in order of materials.

6. The AI-driven full marketing content design method for product packaging planning according to claim 5 is characterized in that: The temporal decay weight function is an exponential decay function, expressed as W(t) = e^(-λt), where t is the time offset and λ is the decay coefficient. The exponential decay function is used to ensure that new materials correspond to high weights and old materials correspond to low weights.

7. The AI-driven full marketing content design method for product packaging planning according to claim 1 is characterized in that: The steps for generating the style unity coefficient include: Receive and parse the dynamic consistency weight matrix to extract the temporal weighted feature scores of each style sub-dimension; Perform weighted fusion or nonlinear merging on the eigenvalues ​​of each column to obtain the initial style uniformity strength value; A normalization algorithm is used to map the initial intensity value to a style unity coefficient between 0 and 1.

8. The AI-driven full marketing content design method for product packaging planning according to claim 1 is characterized in that: The steps of calculating the jump path using the style unity coefficient and dynamically adjusting the jump range based on the tonality unity strength include: When the style unity coefficient exceeds the preset risk threshold, a multi-dimensional style space mapping diagram is constructed based on the brand tone status; Determine the current style coordinates in the style space according to the style unity coefficient and calculate the jump path; Dynamically adjust the jump range based on the strength of brand tonality and unity; Update the style parameters of the content generation model according to the jump coordinates.

Images