Method for automatically generating personalized exploration marketing video of AI digital person

Abstract

The application discloses an AI digital person personalized customized store exploring marketing video automatic generation method, relates to the technical field of artificial intelligence video, and aims to solve the problems of high production cost, low efficiency, digital person homogenization, poor style adaptability, insufficient multi-channel putting adaptation and the like of traditional store exploring videos. The method sets up an enterprise following and shooting double triggering mode by constructing a store exploring exclusive multi-dimensional data input system, relies on an exclusive multi-modal semantic analysis model to disassemble marketing requirements, supports double-path digital person customization of real person replication and modularization self-definition and tone cloning, combines a fine-tuning large model to generate an exclusive marketing script, generates a finished video through a best-selling rhythm replication algorithm and four-dimensional intelligent fusion rendering, and finally outputs after compliance inspection and channel optimization. The application realizes full-process automation, individualization and batch production of store exploring videos, adapts to multi-industry stores and full-platform putting, and greatly reduces production cost and improves marketing effect.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence video technology, specifically to a method for automatically generating personalized store visit marketing videos using AI digital humans. Background Technology

[0002] With the rapid popularization of short video marketing, store visit marketing videos for physical stores in the catering, beauty, retail, and leisure and entertainment industries have become a core means for offline businesses to attract customers and promote their brands. Currently, the production model of store visit marketing videos in the industry is still mainly based on manual shooting, manual editing, and on-camera narration, which has many industry pain points.

[0003] Traditional production methods require the collaboration of professional filmmakers, editors, and real-life store review bloggers. The entire process, from filming and modeling to scriptwriting, voice-over, and editing, involves manual intervention, resulting in long production cycles and high labor costs, making it difficult for small and medium-sized businesses to consistently produce high-quality store review videos. Furthermore, featuring real people in store reviews presents challenges such as scheduling conflicts, inconsistent appearances, limited filming locations, and unstable personas, hindering the standardization and consistent promotion of the store's brand image.

[0004] Existing AI video generation technologies are mostly general-purpose video generation models, lacking specific optimizations for store visit marketing scenarios. They cannot achieve integrated functions such as personalized replication of store-specific digital avatars, accurate replication of viral video styles, precise adaptation of marketing copy, and multi-channel image quality differentiation. Existing technologies generally suffer from severe homogenization of digital avatar images, fixed voices, audio-visual asynchrony, misalignment of video selling points, inability to align with store brand tone, lack of compliance guarantees, and poor adaptability to multi-platform deployment. These shortcomings make it difficult to meet the personalized, standardized, batch, and omni-channel store visit marketing video production needs of physical stores.

[0005] Therefore, this invention proposes an AI-powered digital human marketing video generation method that adapts to specific store visit scenarios, supports full-dimensional customization of real-person digital humans, can replicate corporate brand styles and viral video styles, and is fully automated and adaptable to multi-platform deployment, thus solving many shortcomings of existing technologies. Summary of the Invention

[0006] This invention aims to overcome the problems of high production cost, low efficiency, homogenization of digital humans, poor style adaptability, low audio-visual matching, insufficient adaptation to multi-channel distribution, and weak marketing targeting in existing technologies for store visit marketing videos. It provides an AI-powered method for automatically generating personalized store visit marketing videos, realizing fully automated, personalized, and standardized production of store visit marketing videos from data entry, digital human customization, script generation, video editing and replication, multi-element fusion rendering to compliance quality inspection and optimized distribution across different channels.

[0007] Therefore, this invention provides a method for automatically generating personalized store visit marketing videos using AI digital humans, comprising the following steps: S1: Construct a multi-dimensional data entry system for store visit marketing, obtain customized basic data for store visit marketing and dual-mode one-click shooting trigger commands. The customized basic data for store visit marketing includes structured merchant store information, store real-scene visit scene materials, precise marketing promotion customization needs and scene-adaptive user personalized customization parameters.

[0008] The structured merchant-specific information is a tiered information system adapted to store visit marketing scenarios, including store category tags, core product selling points, differentiated store environment identifiers, time-sensitive promotional activities, and exclusive brand style and tone parameters. The store visit scenario materials are a dedicated library of structured, real-shot materials, including storefront photos, videos of functional areas within the store, and close-up product footage. It also supports uploading real-life photos of faces, bodies, and full-body shots from multiple angles for use in dedicated AI digital human modeling. The customized precise marketing and promotion requirements include multiple... The system includes channel-adaptive promotion parameters, dynamic video duration configuration, contextualized marketing focus weights, and segmented target audience profile tags. The contextualized user-personalized customization parameters include full-dimensional customization parameters for a digital avatar specifically designed for store visits, encompassing live-action modeling to create a customized digital avatar image, live-action voice cloning, customizable broadcast voice, immersive scene broadcast tone, store-adaptive exclusive background music library parameters, and marketing-oriented dynamic subtitle style templates. The dual-mode one-click shooting trigger commands include enterprise brand-specific follow-up shooting mode commands and trending style replication shooting mode commands, serving as the trigger entry point for the entire automated video generation process.

[0009] Furthermore, the one-click follow-up shooting mode is a dedicated follow-up shooting function for enterprise store scenarios. It supports one-click replication of enterprise store shooting angles, scene camera movements, image composition, and brand lens arrangement logic, automatically adapts to the official promotional style of the enterprise, and generates store visit footage that fits the enterprise's brand image.

[0010] Furthermore, in the one-click shooting mode, users can select high-quality popular store visit videos from the platform as templates. AI automatically extracts the same style characteristics of the template videos, such as camera rhythm, transition methods, filter tones, and screen duration allocation, and automatically replicates the shooting and editing logic of the same videos throughout the entire process. At the same time, it can generate video scripts that match the user.

[0011] S2: Construct an AI multimodal semantic parsing model specifically for store visit scenarios. Based on the customized basic data for store visit marketing and the dual-mode one-click shooting trigger command, perform in-depth decomposition and refined feature extraction of scenario-specific needs. Differentiatedly split out the store's exclusive core selling point matrix, the store visit-specific layered video narrative framework, the scenario-matching digital human adaptation feature system including image appearance, cloned timbre, and broadcast voice, the marketing-oriented intelligent video editing rules, and the exclusive style replication features of popular videos.

[0012] S3: Build a refined digital human model library specifically for store visit marketing. Based on the scene-matching digital human adaptation characteristics, match and generate a store-exclusive customized AI digital human with one click.

[0013] This step is the core of digital human customization, possessing three exclusive customization capabilities: real-shot modeling for customized digital human avatars, real-person voice cloning for customized voices, and full-dimensional voice customization. Real-shot modeling for customized digital human avatars can receive real-person frontal, side, and full-body multi-angle real-person video footage uploaded by users. AI 3D reconstruction algorithms are used to achieve high-precision replication of facial features, body proportions, and limb contours. Skin tone, facial details, and hairstyle can be fine-tuned based on the real-person prototype, or customized adjustments can be made using industry-standard models without real-person footage. Real-person voice cloning for customized voices supports the input of short-duration real-person voice samples, completing voiceprint feature extraction and voice cloning training with one click to generate a highly realistic, personalized narration voice with the quality, speed, and emotion of a real person speaking. Full-dimensional voice customization, in addition to cloning voices, supports the free selection of a vast library of built-in commercial voices, allowing users to independently adjust the broadcast speed, tone, emotional intensity, and pause rhythm to match the atmosphere of different store visits in industries such as catering, beauty, and retail.

[0014] Simultaneously, a three-in-one linkage generation mechanism is constructed, integrating shooting mode, store tone, and core selling points. By fine-tuning the exclusive training weight of the large language model for store visits, exclusive store visit marketing copy and segmented video narration scripts are automatically generated, which are adapted to the corporate brand style, the rhythm of shooting the same popular products, and the core selling points and brand tone of the store.

[0015] Furthermore, the dedicated refined digital human model library for store visit marketing includes built-in basic digital human models adapted to all store visit industries, such as catering, beauty, retail, and leisure and entertainment. It supports two core digital human customization paths: Path 1 is real-life replica customization, which relies on user-uploaded multi-angle real-life images to complete 3D reconstruction, replicating the real person's facial features, body shape, and expression in a 1:1 ratio, generating a unique store digital human; Path 2 is modular customization, which does not require real-life shooting, and directly fine-tunes the digital human's skin tone, intelligently replaces industry-adaptive clothing, and customizes the exclusive posture for store visit broadcasts. Both customization paths are equipped with voice cloning and custom broadcast voice functions, which can match exclusive voice to immersive store visit broadcasts, realizing two-way, full-dimensional personalized customization of digital human image and voice, creating a unique digital human image for each store.

[0016] Furthermore, when the finely tuned large language model generates exclusive store visit marketing copy and video narration scripts, it constructs a three-dimensional matching mechanism of audience profile, marketing focus, and video duration. It dynamically adapts the copywriting marketing style according to the segmented target audience and the store's marketing focus, while also adapting to multiple video formats such as short videos, long videos, and live stream clips. This enables modular, segmented, and differentiated precise writing of scripts, which aligns with the conversion needs of store visit marketing. After generating the narration text, it can be directly bound to a cloned real person's exclusive voice, generating exclusive narration audio that matches the digital human image with one click.

[0017] S4. Based on the exclusive layered video narrative framework for store visits, marketing-oriented intelligent video editing rules, and exclusive style replication features of viral videos, an intelligent visual rhythm replication algorithm for viral videos is constructed. This algorithm performs intelligent layered editing, scene-logical image splicing, adaptive adaptation of store-style exclusive filters, and exclusive transition timing processing for viral videos. It accurately replicates the shot rhythm, visual logic, and playback rhythm of viral videos, generating standardized and differentiated exclusive store visit video clips.

[0018] S5. Construct a four-dimensional intelligent fusion rendering system integrating digital human, materials, copywriting, and audio. This system integrates a customized AI digital human image generated through real-shot modeling and voice cloning, exclusive cloned narration audio, store-specific marketing copywriting, and standardized store visit video clips into a unified intelligent fusion. It also links and matches store-specific background music parameters and marketing dynamic subtitle styles to achieve adaptive beautification rendering of all elements. This automatically generates a complete AI digital human store visit marketing video with a unified style and prominent selling points. During the rendering process, the digital human's lip movements and cloned narration audio are synchronized at the millisecond level to ensure that the lip movements of the real-person cloned digital human naturally match the customized voice.

[0019] Furthermore, the four-dimensional intelligent fusion rendering system includes high-precision digital human cutout synthesis for store visits, intelligent adaptive layout of marketing copy and subtitles, millisecond-level synchronous alignment of cloned narration audio with the mouth shape of the customized digital human in real-shot footage, and dynamic intelligent adjustment of background music volume according to the scene. This solves the problems of fragmented images, audio-visual asynchrony, stiff matching of digital human audio and visuals, and misalignment of selling points in traditional synthesis, and achieves integrated collaborative rendering of real-shot modeled digital humans, store real-shot materials, marketing copy, cloned narration audio, and background music.

[0020] S6. Establish a dedicated compliance quality inspection and channel-specific optimization system for store visit marketing videos. Perform scenario-based compliance testing and platform-differentiated image quality retouching on finished videos. Accurately remove sensitive, infringing, and illegal content in store visit scenes, correct original flaws in store footage, match the exclusive image quality delivery standards of each short video platform, and finally output personalized customized store visit marketing videos that can be directly delivered across all channels.

[0021] Furthermore, the aforementioned channel-specific image quality enhancement system constructs an independent image quality parameter adaptation model for existing mainstream marketing platforms. It supports differentiated optimizations such as intelligent matching of screen resolution, accurate color correction of store-shot videos, dedicated removal of noise in materials, intelligent repair of handheld shooting with anti-shake, and enhancement of product detail clarity, adapting to the image quality standards recommended by the algorithms of each platform. Simultaneously, it performs corresponding platform-specific portrait image quality enhancement processing on the digital human images generated by real-shot modeling to ensure the clarity of cloned digital human images during projection.

[0022] Furthermore, this method constructs a full-link intelligent distribution system for store visit videos. The generated personalized store visit marketing videos support one-click standardized export, one-click publishing with cross-platform permissions, and are equipped with online real-time preview, targeted secondary editing of marketing parameters, digital human image reshoot and replication, re-uploading and cloning of voice samples, readjustment of vocal parameters, and custom parameter reset and regeneration functions, realizing a fully automated closed loop of store visit video production, digital human image voice iteration, optimization, and distribution.

[0023] Compared to existing technologies, the advantages of this invention are: This invention constructs a multi-dimensional data entry system specifically for store visit marketing, enabling standardized and refined entry of store information, scene materials, marketing needs, and personalized parameters. Coupled with dual-mode shooting trigger commands, it can be adapted to two core marketing scenarios: standardized corporate brand promotion and viral video traffic generation, making it more adaptable and targeted.

[0024] This invention builds a dedicated AI multimodal semantic parsing model that can deeply analyze the needs of store visit marketing, accurately extract store selling points, video narrative logic, digital human adaptation features, and popular style features, solving the problem that general AI models cannot fit the segmented scenarios of store visits and the marketing focus is vague.

[0025] This invention innovatively sets up a dual-path digital human customization system, supporting 1:1 three-dimensional replication of real people and modular customization, combined with real person voice cloning and full-dimensional voice adjustment functions, to achieve exclusive and personalized customization of digital human image and voice, completely solving the defects of traditional digital human serious homogenization and fixed character design, and creating a store's exclusive IP image.

[0026] This invention utilizes a large language model with customized store visit weighting and a three-dimensional matching mechanism to adaptively generate differentiated marketing copy and segmented narration scripts based on the audience, marketing focus, and video length. It adapts to various video scenarios, precisely meets marketing conversion needs, and achieves a higher marketing conversion rate compared to general copy generation technologies.

[0027] This invention constructs an intelligent algorithm for replicating the visual rhythm of videos, which can accurately replicate the rhythm, transitions, filters, and duration logic of hit videos, and automatically complete the editing and splicing of store materials without manual editing, greatly reducing the threshold and cost of video production and enhancing the potential of videos to become hits.

[0028] This invention features a unique four-dimensional intelligent fusion rendering system that enables integrated collaborative rendering of digital humans, live-action footage, text, and audio. It achieves millisecond-level lip-sync, adaptive subtitle layout, and dynamic music adjustment, solving the technical pain points of traditional composite videos such as fragmented visuals, audio-visual asynchrony, and misaligned selling points, resulting in a significant improvement in the quality of the final video.

[0029] This invention establishes a dedicated compliance quality inspection and channel-specific image quality retouching system, which can automatically remove illegal content, repair flaws in actual shooting, adapt to the image quality standards of various mainstream short video platforms, and ensure that videos are compliant and can be deployed with strong traffic adaptability.

[0030] This invention constructs a fully automated closed-loop system that supports one-click video export and distribution, secondary parameter editing, and iterative optimization of digital human avatar voice, realizing full automation of video production, optimization, and distribution, and significantly improving the marketing and operational efficiency of merchants. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of the overall method flow provided in the embodiments of the present invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0034] The present invention provides the following embodiments: First Implementation Example: Customized Implementation Example of Corporate Brand Exclusive Follow-up Photography Mode Implementation Scenarios Addressing the official marketing needs of chain restaurant brands, this invention leverages the S1-S6 full-process technology system and employs a brand-specific follow-up shooting mode. It replicates the brand's unified shooting style and composition, customizes a brand-specific AI digital avatar, and generates standardized store visit marketing videos that align with the company's system and brand tone. These videos are suitable for chain store brand promotion and daily operation and placement of official accounts. This example selects a domestic chain tea beverage brand (200+ offline stores) for single-store implementation testing.

[0035] Core parameter configuration

[0036] Based on a multi-dimensional data entry system, standardized brand-specific data is entered, and the enterprise follow-up shooting mode trigger command is locked. The specific entry parameters and configuration standards are shown in the table below:

[0037] Core Implementation Process

[0038] Step 1: Input the aforementioned brand-specific structured data, store footage, and live-action modeling materials to trigger the enterprise's brand-specific follow-up shooting mode command; Step 2: Use an AI multimodal semantic analysis model to deconstruct the brand's core selling points and standardized video narrative framework, extracting brand-specific camera movement, composition, and screen layout features; Step 3: Generate a store-specific digital human through the live-action model replication path, complete the cloning of the store manager's voice, and generate a 60-second brand-specific narration script based on the fine-tuned large model, with a copywriting style that matches the brand's minimalist and youthful tone; Step 4: According to the brand's official editing rules, perform layered editing and image splicing of the store materials, adapt to the brand's exclusive filter parameters, and unify the brand's visual style; Step 5: Activate the four-dimensional intelligent fusion rendering system to achieve millisecond-level synchronization between the digital human's lip movements and the narration audio, adapting to the brand's BGM and standardized subtitle styles; Step 6: Complete compliance quality inspection and channel-specific image quality refinement, remove material noise, correct color deviations, and adapt to the 1080P delivery standards of various platforms.

[0039] 4. Implementation Results

[0040] The final product is a standardized 60-second store visit video with a unified visual style, narration, and digital avatar across all stores, ensuring a seamless brand experience. The audio-visual synchronization error is ≤20ms, the compliance rate is 100%, and the generation time for a single video is ≤8 minutes. Compared to traditional live-action shooting and editing, the efficiency is improved by 85%, perfectly meeting the standardized marketing needs of chain brands.

[0041] Example 2: Customization Example of a Popular Style Replica Model

[0042] Implementation Scenarios

[0043] To address the need for small and medium-sized beauty stores to expand their reach, this method requires no professional shooting and editing skills. It selects high-traffic, trending store visit videos from the platform as templates and adopts a one-click shooting mode to replicate the trending videos. AI replicates the rhythm, transitions, and filter styles of the trending videos and combines them with the store's own selling points to generate differentiated, high-traffic store visit videos. This method is suitable for natural traffic generation on short video platforms and for local store visits and product seeding. This example focuses on the implementation and testing of community beauty stores (nail and eyelash salons).

[0044] The core parameter configurations are as follows:

[0045] Core Implementation Process

[0046] First, we input exclusive information about beauty salons, service materials, and data on the profiles of young women in the same city, upload a popular template video, and trigger the same-style replication mode. Second, an AI semantic analysis model extracts all visual, rhythmic, and textual structural features of the template, breaks down the core selling points of the salon, and constructs a layered narrative framework adapted to a 45-second short video. Relying on a refined digital human model library, we use a modular customization approach to generate a fresh and clean style digital human adapted to the beauty industry. Through a three-dimensional matching mechanism of audience, marketing focus, and duration, we generate a salon-exclusive seeding script with a popular structure. Using a smart algorithm for replicating the visual rhythm of the same-style video, we accurately replicate the template's timing, transitions, and filter rhythms, replacing the salon's actual footage. We complete the four-dimensional fusion rendering of digital human, text, audio, and materials, and optimize the dynamic pop-up style of subtitles. Finally, we perform compliance quality checks, remove sensitive wording that is subject to platform traffic restrictions, and adapt to short video traffic push standards.

[0047] Implementation effect

[0048] The generated store visit videos perfectly replicate the rhythm of viral hits, with precise timing and a consistent visual style. They also perfectly match the store's own service selling points, avoiding the problem of homogenized copying. The video completion rate is 42% higher than traditional self-made videos, and the local exposure is 38% higher. No manual editing or layout is required, achieving the dual effects of one-click replication of viral hit style and store-specific customization.

[0049] Example 3: Industry-wide Modular Digital Human Customization in All Dimensions

[0050] Implementation Scenarios

[0051] Targeting four major store visit industries—catering, beauty, retail, and leisure and entertainment—this invention eliminates the need for real-life footage. Employing a modular, customizable path and industry-specific parameters, it enables comprehensive customization of the digital human's appearance, voice, posture, and delivery style, adapting to the marketing tone of different industry stores and validating its multi-industry universal adaptability.

[0052] The following is a table of differentiated parameters for customized digital humans across multiple industries:

[0053] Core Implementation Process

[0054] For stores in different industries, corresponding category tags, marketing needs, and audience profile parameters are entered; without the need for real-life footage, the system calls upon the corresponding industry-specific digital human models from the model library to complete skin tone adjustments, clothing replacements, and posture customization; it matches industry-specific voices and broadcast parameters, and customizes speech rate, tone, and emotional weight; through fine-tuning the weights of the large model, it generates segmented narration scripts adapted to the industry tone and audience needs; combined with real-life store footage, it matches industry-specific filters, transitions, and background music to complete four-dimensional intelligent fusion rendering; it performs compliance optimizations based on industry characteristics, avoiding food safety violations in the catering industry and exaggerated claims about medical efficacy in the beauty industry, ultimately generating customized store visit videos for each industry.

[0055] Implementation effect

[0056] All four major industries have achieved rapid, fully modular customization of exclusive digital avatars without real-person filming. The avatars' appearance, voice, and broadcasting style are highly compatible with the industry, creating a seamless experience. The digital avatar customization success rate is 100%, and the generation time for customized videos in a single industry is ≤10 minutes. The avatars are suitable for all categories of store visit marketing needs, solving the pain points of traditional digital avatars' single style and poor industry adaptability.

[0057] In summary, the above description is merely a preferred embodiment of the technical solution of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for automatically generating personalized store visit marketing videos using AI digital humans, characterized in that... include: S1: Construct a multi-dimensional data entry system for store visit marketing, obtain customized basic data for store visit marketing and dual-mode one-click shooting trigger instructions. The customized basic data for store visit marketing includes structured merchant store information, store real-scene visit scene materials, precise marketing promotion customization needs and scenario-adaptive user personalized customization parameters. The structured merchant-specific information is a tiered information system adapted to store visit marketing scenarios, including store category tags, core product selling points, differentiated store environment identifiers, time-sensitive promotional activities, and exclusive brand style and tone parameters. The store visit scenario materials are a dedicated library of structured, real-shot materials, including storefront photos, videos of functional areas within the store, and close-up product footage. It also supports uploading real-life photos of faces, bodies, and full-body shots from multiple angles for use in dedicated AI. The system includes digital human replica modeling; the customized marketing and promotion requirements include multi-channel adaptation parameters, dynamic video duration configuration, scenario-based marketing focus weights, and segmented target audience profile tags; the scenario-adaptive user personalization parameters include full-dimensional customization parameters for store visits, covering real-life shooting modeling to customize the digital human image, real-life voice cloning, custom broadcast voice, immersive scene broadcast tone, store-adaptive exclusive background music library parameters, and marketing-oriented dynamic subtitle style templates; the dual-mode one-click shooting trigger commands include enterprise brand exclusive follow-up shooting mode commands and popular style replica shooting mode commands, which serve as the trigger entry point for the entire automated video generation process. S2: Construct an AI multimodal semantic parsing model specifically for store visit scenarios. Based on the customized basic data for store visit marketing and the dual-mode one-click shooting trigger command, perform in-depth decomposition and refined feature extraction of scenario-specific needs. Differentiatedly split out the store's exclusive core selling point matrix, the store visit-specific layered video narrative framework, the scenario-matching digital human adaptation feature system including image appearance, cloned timbre, and broadcast voice, the marketing-oriented intelligent video editing rules, and the exclusive style replication features of viral videos. S3: Build a refined digital human model library specifically for store visit marketing, and generate a customized AI digital human for the store with one click based on the scene matching digital human adaptation characteristics. This step is the core of digital human customization, possessing three exclusive customization capabilities: Real-world modeling and customization of digital human images: Receive real-world images uploaded by users from multiple angles, including front, side, and full-body shots. Use AI 3D reconstruction algorithms to accurately replicate facial features, body proportions, and limb contours. Skin tone, facial details, and hairstyle can be fine-tuned based on the real-world prototype. Alternatively, users can choose to customize adjustments using an industry-standard model without relying on real-world footage. Real-person voice cloning for customized voice: Supports recording short-duration real-person voice samples, one-click completion of voiceprint feature extraction and voice cloning training, generating exclusive narration voice that highly restores the texture, speed and emotion of real people speaking. Full-dimensional customizable sound configuration: In addition to cloned voices, it supports the free selection of a large number of commercial voices built into the system, and allows users to adjust the broadcast speed, tone, emotional intensity, and pause rhythm to match the atmosphere of different stores such as catering, beauty, and retail. Simultaneously, a three-in-one linkage generation mechanism is constructed, integrating shooting mode, store tone, and core selling points. By fine-tuning the exclusive training weights of the large language model for store visits, exclusive store visit marketing copy and segmented video narration scripts are automatically generated, which are adapted to the corporate brand style, the rhythm of shooting the same popular products, and the core selling points and brand tone of the store. S4. Based on the exclusive layered video narrative framework for store visits, marketing-oriented intelligent video editing rules, and exclusive style replication features of viral videos, an intelligent visual rhythm replication algorithm for viral videos is constructed. This algorithm performs intelligent layered editing, scene-logical image splicing, adaptive adaptation of store-style exclusive filters, and exclusive transition timing processing for viral videos. It accurately replicates the shot rhythm, visual logic, and playback rhythm of viral videos, generating standardized and differentiated exclusive store visit video clips. S5. Construct a four-dimensional intelligent fusion rendering system integrating digital human, materials, copywriting, and audio. This system integrates a customized AI digital human image generated through real-shot modeling and voice cloning, exclusive cloned narration audio, store-specific marketing copy, and standardized store visit video clips. It also links and matches store-specific background music parameters and marketing dynamic subtitle styles to achieve full-element adaptive beautification rendering. This automatically generates a complete AI digital human store visit marketing video with a unified style and prominent selling points. During the rendering process, the digital human's lip movements and cloned narration audio are synchronized at the millisecond level to ensure that the lip movements of the real-person cloned digital human naturally match the customized voice. S6. Establish a dedicated compliance quality inspection and channel-specific optimization system for store visit marketing videos. Perform scenario-based compliance testing and platform-differentiated image quality retouching on finished videos. Accurately remove sensitive, infringing, and illegal content in store visit scenes, correct original flaws in store footage, match the exclusive image quality delivery standards of each short video platform, and finally output personalized customized store visit marketing videos that can be directly delivered across all channels.

2. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, The one-click follow-up shooting mode mentioned in step S1 is a dedicated follow-up shooting function for enterprise store scenarios. It supports one-click replication of enterprise store shooting angles, scene camera movements, image composition, and brand lens arrangement logic, automatically adapts to the official promotional style of the enterprise, and generates store visit footage that fits the enterprise's brand image.

3. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, In step S1, the one-click shooting mode allows users to select popular store visit videos from the platform as templates. AI automatically extracts the same style characteristics of the template videos, such as camera rhythm, transition methods, filter tones, and screen duration allocation. The entire process of automatically replicating the shooting and editing logic of the same videos is automated. At the same time, it can generate video scripts that match the user.

4. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, In step S3, the refined digital human model library specifically designed for store visit marketing includes built-in basic digital human models adapted to the entire store visit industry, including catering, beauty, retail, and leisure and entertainment. It primarily supports two digital human customization paths: Path 1 is real-life replica customization, which relies on user-uploaded multi-angle real-life images to complete 3D reconstruction, replicating the facial features, body shape, and expression of a real person at a 1:1 ratio, generating a unique store digital human; Path 2 is modular customization, which does not require real-life photography, directly fine-tuning the digital human's skin tone, intelligently replacing industry-appropriate clothing, and customizing the exclusive posture for store visit broadcasts. Both customization paths are equipped with voice cloning and custom broadcast voice functions, which can match exclusive voices for immersive store visit broadcasts, achieving two-way, full-dimensional personalized customization of the digital human's image and voice, creating a unique digital human image exclusive to each store.

5. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, In step S3, when the finely tuned large language model generates exclusive store visit marketing copy and video narration script, a three-dimensional matching mechanism of audience profile, marketing focus, and video duration is constructed. The copywriting marketing style is dynamically adapted according to the segmented target audience and the store's marketing focus. At the same time, it adapts to multiple video scenarios such as short videos, long videos, and live broadcast segments, realizing modular, segmented, and differentiated precise writing of scripts to meet the conversion needs of store visit marketing. After generating the narration text, it can be directly bound to the cloned real person's exclusive voice, generating exclusive narration audio that matches the digital human image with one click.

6. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, The four-dimensional intelligent fusion rendering system includes high-precision digital human cutout and compositing for store visits, intelligent adaptive layout of marketing copy and subtitles, millisecond-level synchronization and alignment of cloned narration audio with the mouth shape of the customized digital human in real-shot footage, and dynamic intelligent adjustment of background music volume according to the scene. It solves the problems of fragmented images, audio-visual asynchrony, stiff matching of digital human audio and visuals, and misalignment of selling points in traditional compositing, and realizes integrated collaborative rendering of real-shot modeled digital humans, store real-shot materials, marketing copy, cloned narration audio, and background music.

7. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, In step S6, the channel-specific image quality enhancement system constructs an independent image quality parameter adaptation model for existing mainstream marketing platforms. It supports differentiated optimization such as intelligent matching of screen resolution, accurate color correction of store real-shots, special removal of material noise, intelligent repair of handheld shooting anti-shake, and enhancement of product detail clarity, and adapts to the image quality standards recommended by the algorithms of each platform. At the same time, it performs corresponding platform portrait image quality enhancement processing on the digital human images generated by real-shot modeling to ensure the clarity of cloned digital human images for projection.

8. The method for automatically generating personalized store visit marketing videos using AI digital humans according to claim 1, characterized in that, This method constructs a full-link intelligent distribution system for store visit videos. The generated personalized store visit marketing videos support one-click standardized export, one-click publishing with cross-platform permissions, and are equipped with online real-time preview, targeted secondary editing of marketing parameters, digital human image reshoot and replication, re-uploading and cloning of voice samples, readjustment of vocal parameters, and custom parameter reset and regeneration functions. This realizes a fully automated closed loop for the entire process of store visit video production, digital human image voice iteration, optimization, and distribution.

Images