Poster generation method and device based on RAG poster design and electronic equipment

Abstract

The invention provides an RAG poster design generation method and device and electronic equipment, the method converts creativity into reality, and firstly, poster visual design content element information input by a user in a first area of a first interface of an RAG poster making tool is obtained; then, an RAG intelligent agent identifies the input poster visual design content element information and understands semantics of the poster visual design content element information, and a ReAct mode is used as a starting point of an RAG intelligent agent design mode; poster visual design generation background elements and automatic typesetting design parameters matched with the news broadcast and poster visual design content element information are obtained, and a target poster generated based on the poster visual design generation background and the automatic typesetting design parameters is displayed in a second area of the first interface.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a poster generation method, device and electronic device based on RAG poster design. Background Art

[0002] With the continuous development of artificial intelligence technology, we can use LLM (Large Language Model), text-image model, etc. to write manuscripts or generate images.

[0003] However, since the underlying principle of the LLM model is to make predictions based on mathematical probability, its model output is essentially a result of probabilistic prediction, so the LLM model sometimes speaks nonsense or generates some seemingly plausible answers. In areas where the model is not good at, the hallucination problem will be more serious; especially compared with the Internet industry, professional knowledge samples in the media field are scarce and difficult to accumulate. Therefore, the LLM model cannot meet the needs of designers when automatically generating poster-related content. The text-based image model can only generate images that match the description, and can only simply implement poster layout design, and cannot adapt the poster content from the text (it cannot be commercially used).

[0004] Moreover, currently most of the writing or images are realized based on web pages, and it is very inconvenient for users to operate through command lines. Summary of the invention

[0005] In order to solve or partially solve the problems existing in the related art, this application provides a method, device and electronic device for automatically designing RAG posters. RAG can not only generate content according to templates, but also combine the team's actual work style and visual language to provide truly valuable output. It can facilitate user operation while generating posters that meet the user's commercial needs for understanding posters and protect the user's training data from being leaked.

[0006] The first aspect of the present application provides a poster generation method based on RAG poster design, comprising: The RAG agent uses the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, recognizes the input poster visual design content element information and understands the semantics of the poster design content element, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains the poster design generation background and automatic typesetting design parameters that match the poster design content element information (the overall framework of the poster design makes people feel more comfortable, the matching is more fluent, the proportion is better, it will not be too divided, the high-frequency elements have a higher priority, a larger display space, and a more eye-catching color scheme. In fact, the golden ratio, graphic design rules, etc. are considered when designing modular poster size element information); Based on the poster visual design generating background elements and the automatic typesetting design parameters, the target poster output by the poster design generating model is displayed in the second area of ​​the first interface (using the F-type browsing mode proposed by Nielsen (because the browsing mode is like the English letter F, it is called the F mode), that is, the reader reads from left to right, then scans vertically, and continues to read from left to right after locating the point of interest. This dynamic line conforms to the natural reading mode of humans and greatly improves the scanning efficiency); Among them, the RAG agent is obtained after training based on the personalized vector database configured in the private domain container as a training set and the basic LLM large model and basic literary graph model based on the public domain. The personalized vector database at least stores the promotional poster design knowledge template that conforms to the corporate culture content of the user. The RAG agent can at least reason and act to solve the language reasoning and decision-making tasks of diversified poster design.

[0007] The second aspect of the present application provides another poster generation method based on RAG poster design, comprising: The RAG agent uses the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, recognizes the input poster visual design content element information and understands the semantics of the poster design content element, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains a poster design generation background that matches the poster design content element information; The manual design operation of the user using the poster visual design content element information within the poster design generation background is received to obtain a target commercial poster.

[0008] A third aspect of the present application provides a poster generation device based on RAG poster design, comprising: An acquisition module is used to obtain the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, and the RAG agent identifies the input poster visual design content element information and understands the semantics of the poster design content element, using the ReAct mode (starting from ReAct, there are two routes: one focuses on the planning ability of the Agent, including REWOO, Plan&Execute, and LLM Compiler. The other focuses more on reflection ability, including Basic Reflection, Reflexion, Self Discovery, and LATS. By combining reasoning and acting in the language model to solve a variety of language reasoning and decision-making tasks) as the starting point of the RAG agent design mode, and obtains the poster visual design generation background elements and automatic typesetting design parameters that match the poster design content element information; A generation module generates background elements (matching content elements, integrating visual art to attract readers) and the automatic typesetting design parameters based on the poster visual design, and displays the target commercial poster output by the poster design generation model in the second area of ​​the first interface; wherein the RAG agent is obtained after training based on the personalized vector database configured in the private domain container as a training set and the basic LLM large model and basic literary graph model based on the public domain, and the personalized vector database at least stores a promotional poster design knowledge template that conforms to the corporate culture content of the user, and the RAG agent can at least reason and act to solve the language reasoning and decision-making tasks of diversified poster design.

[0009] There is also the combination of design core values ​​and RAG technology. By simplifying complex design principles, optimizing the design-to-code conversion process, pragmatically solving practical problems, and moving towards "collaborative creation", RAG tools can truly unleash the potential of designers. In this process, the role of designers is not only not replaced by RAG, but becomes more important, because they are always the source of creativity and the key figures in transforming technology into user experience.

[0010] The fourth aspect of the present application provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the RAG poster generation method as described above.

[0011] The technical solution provided by this application may have the following beneficial effects: The present application provides a poster generation method, device and electronic device based on RAG poster design. The method first obtains the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster production tool; then the RAG agent recognizes the input poster visual design content element information and understands the semantics of the poster visual design content element information, uses the ReAct mode as the starting point of the RAG agent design mode, obtains the poster design generation background and automatic typesetting design parameters that match the poster visual design content element information, and displays the target poster generated based on the poster design generation background and automatic typesetting design parameters in the second area of ​​the first interface; it can be seen that in the embodiment of the present application, due to the use of a self-developed special software interface, it can meet the user's operating habits and facilitate the user's operation, and because the private domain container can protect the personalized database from leakage and the agent can reason and act, thereby generating a target poster that meets the user's needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The above and other objects, features and advantages of the present application will become more apparent by describing in more detail exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.

[0013] Figure 1 It is a flowchart of a poster generation method based on RAG poster design disclosed in an embodiment of the present application; Figure 2 is a schematic diagram of the display effect of the first interface shown in an embodiment of the present application; Figure 3 is another display effect schematic diagram of the first interface shown in the embodiment of the present application; Figure 4 is a schematic diagram of the display effect of the modification menu in the second area of ​​the first interface in an embodiment of the present invention; Figure 5 is another flowchart of a poster generation method based on RAG poster design shown in an embodiment of the present application; Figure 6 It is a structural schematic diagram of a poster generation device based on RAG poster design disclosed in an embodiment of the present application; Figure 7 It is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present application. DETAILED DESCRIPTION

[0014] The embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.

[0015] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0016] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0017] How to improve the readability and usability of poster layout elements through RAG poster design upgrades, and explore strategies for mobile terminal adaptation, so that poster layout element data can also tell a clear story in a small space. In addition to the illusion problem, the large language model LLM (Large Language Model) gives specious answers to areas that it is not good at. It is itself a black box. Users have no way of knowing what data this model uses for training and what the alignment strategy is. Therefore, it is even more difficult to trace the answers generated by the large model. Based on this, the RAG intelligent body of the embodiment of the present application plugs in the information poster elements and information poster templates in the personalized vector database, limits the illusion problem of LLM, and trains the poster design generation model through a private domain container. The private domain container has a personalized professional personality vector database, which stores information posters, so that the intelligent body can master the core skills of graphic visual design, including plane composition, color composition, pattern modeling, font design, logo design, etc. These skills are the foundation of graphic visual design majors. By completing relevant learning and practice, they can fully demonstrate their professional abilities and achievements in the field of graphic visual design.

[0018] The embodiment of the present invention considers two types: manual and automatic. The information poster elements and information poster templates are stored in the aforementioned database for training. The database is a distributed vector database and is positioned as a tool-type product. Tool products are used to help users solve specific problems. They have very clear functions and uses. Users use them to improve efficiency (in this era of information explosion, people are unwilling to waste a second on meaningless things, and the same is true for posters. If they don’t recognize them after taking a look, they may just swipe away. Therefore, in the design of posters, which are a layout that carries traffic, it is very important to highlight the value of users), produce content, etc. to solve very specific problems (grasp the readers’ attention), The most critical point for readers is to like it. In order to convey the value of the poster, the design constantly pricks the readers' defenses, breaks the readers' defenses, and then amplifies them. In fact, most readers are not rational consumers. Because of human nature, our way of thinking is mostly dominated by systems, unconsciousness, speed, and brainless. There are many inducements that make readers lose their rationality, such as: like, miss, fear, curiosity, owe, vanity, rebellion, etc. Among them, like and miss are the most commonly used in design. Appropriate use of principles and techniques, increase the sense of urgency and scarcity elements, create a feeling of loss if missed, and let readers press the "reading button" with absolute determination when they hesitate. From the perspective of the readers' goal of using reading posters, tool-type products are problem-solving oriented, strong purpose, used to solve specific problems, and strong function-driven. Therefore, another feature of tool-type products is that they are very purposeful and have clear milestones. For example, we go to the hospital only when we feel uncomfortable in daily life, and we go to the gas station only when the car is out of gas. After going there, we hope to solve the corresponding problems, and the milestones to achieve the goals are very clear. But in daily life, people may go shopping and chat with people whether they have something to do or not. These actions are not driven by a strong purpose, and there are no clear milestones. For example, whether or not you buy something does not affect shopping. Tool-type products, like hospitals and gas stations, are designed to solve specific problem scenarios. They are highly purposeful and have clear milestones. From the perspective of product goals, productivity tools help users improve productivity and allow users to complete content production more efficiently and simply. From the perspective of interactive form, users construct content in a canvas through various operations, such as a poster. Just like a sculptor, he constantly carves and polishes around an object, and finally produces a complete poster design object. The information poster template library vector data training, that is, using tools to support creative activities well, adopts the design that triggers the user's "flow" state. This is a state of full involvement. During the process, people will ignore the passage of time and other distractions, allowing people to continue working for several hours. On the contrary, bad tools will distract people. For example, when a user needs to wait half a minute to see the incremental results of her modifications, her attention will be distracted and her flow state will be interrupted. Continuous bad interactive experience will greatly challenge the user's patience.Therefore, for tool software design, how to make users focus on creative activities, achieve the state of "flow", and immerse users in it is an important goal of training software design. Incrementally constructed interaction, that is, the user operation path is around an object in a canvas, calling a series of operations around (canvas) (left and right side functions) to act on this object. Compared with browsing and control products, the interaction has a very high frequency. It is precisely because of this interactive form that the interaction of productivity tools usually has very high requirements. In order to make users operate smoothly, produce efficiently, and produce good results more simply, the design of productivity tool products is often more challenging. These interactive forms are born to solve specific problems and have been precipitated as classic solutions to solve such problems. Productivity tools are born for content production. From a time perspective, classic productivity tools are stable in framework. From the perspective of interactive experience, provide content producers with efficient and easy-to-use operating experience. For example, the role-based interface allows the software to continue to expand on the basis of the original framework when it is facing different markets, and the rules are applicable to multiple scenarios, making addition simpler and easier.

[0019] The method of the embodiment of the present application can be implemented in the form of self-developed software running in the operating system. The goal of the productivity tool software is to allow users to efficiently produce good content. During the use of the software, it has high-frequency interactive features, so users will pursue extreme operating efficiency, just like the setting of each process of the assembly line. How to enable workers to complete various operations most smoothly and efficiently is a link that greatly affects work efficiency. The most fundamental means to improve efficiency is the changes brought about by technological progress. After all, "no matter how reasonable the assembly line process management is, it will never be better than production automation", but the subversion of technology is not achieved overnight, so we will not discuss how to improve the efficiency of productivity tools from the perspective of technical means, but how to improve the interface operation efficiency of productivity software on the market from the perspective of product design. The stages of the user-product interaction process are divided into input, adjustment, viewing and output operations. The focus of poster design tool products in these three stages is different: combining the poster design characteristics to maximize the efficiency of the core stage interaction operation can impress users. The following will focus on how to design efficient interactions around the input, adjustment and viewing stages.

[0020] Input efficiency: Placing points, lines, surfaces, bodies, element templates, and even importing them into the canvas are all input processes. How to make the input process more efficient? To make input more efficient, there are several ways to improve efficiency from the object itself, space, and time: From the perspective of the granularity of input objects, the larger the granularity of input objects, the higher the efficiency. For example, based on a template and component, prefabricated content and templates. From the perspective of the trigger path, a transient interactive form is provided to allow users to access it more efficiently and instantly.

[0021] For example, in the material library, the ability to provide by sub-category is provided so that users can quickly access it. From the timing of triggering input, the content that users want at the current stage can be predicted, so that users can obtain and input it faster. For example, in the background manual design interface, some element options that match the style are actively popped up, so that users can place the appropriate model into the scene faster.

[0022] Adjustment efficiency: After input, some properties will be adjusted, such as color, position, size, angle, and mouse click, selection, drag, pull, and other interactions are all adjustment operations, which are also the most frequent operation types of complex self-developed intelligent visual design editor software. Input operations usually improve efficiency from trigger to object, and here are two points: the trigger is close enough and the adjustment is fast enough.

[0023] Trigger close enough: In high-frequency operation interfaces, being physically close enough is a tried-and-tested method. This case designs a contextual menu based on object selection, which is a good solution to the problem of invariant complex menu calls. For example, we added an explicit contextual menu, the purpose of which is to allow the mouse to operate immediately after selecting an object.

[0024] Adjustment is fast enough: In addition to making the controller and the controlled object physically close enough, the object can also be adjusted faster by providing a variety of quick adjustment capabilities. For example, shortcut key calls, such as in design tools, involve a large number of material brushes. If you call the operation in the menu every time, the efficiency is very low, so users rely heavily on this shortcut key to improve efficiency. Of course, there are many such shortcut operations. Repeat the previous step or even several steps: In some operations with repeated operations, the software usually provides quick operations to repeat the previous step or even several steps. Efficiency of viewing operations. Viewing mainly refers to viewing the constructed content objects. During the editing process, it is usually accompanied by input and adjustment. In order to allow users to view efficiently when editing content, the main performance of various software is "more and more effective acquisition of useful information". There are several types of processing methods in "space" and "time".

[0025] In terms of space: the canvas content area is "big enough". Since the goal of productivity tools is to produce content, it is very important to ensure the viewing efficiency of the content. The main content should be given priority in content display, and the canvas where the object is located should be large enough to ensure basic viewing efficiency. For content outside the content area, due to limited space and in order to ensure the efficiency of operation, the operation commands will appear in a very compact state. The essence is that the canvas space needs to give way to the canvas display area.

[0026] In terms of time: multiple lines in parallel and multiple task views are used to improve the viewing efficiency of the software. In order to make the interface spacious enough for both the material viewing and editing areas, modes such as displaying element materials, backgrounds, templates, layers, etc. on the left and right sides are provided to greatly improve the efficiency of material viewing. Confirm the position through multiple views to achieve the effect of accurate editing while watching. In addition to the space and time mentioned above, exclusive and convenient functions are also provided for viewing, such as hiding, managing and viewing objects based on layers or object properties. The more functions a software has, the more complex it is, and the more complex it is, the slower the user will get started, which seems to be an unsolvable dead loop. However, our design approach draws on the predecessors to reduce the resistance to getting started with the product (simply pop up according to needs).

[0027] Regarding the systematic consistency experience: One of the main benefits of consistent user experience in the poster design product in the embodiment of the present application is skill transfer. Consistency helps skill transfer. When the user has learned A, when the new function B appears, he may get started quickly. This is the biggest benefit of consistency. From appearance, behavior, to cognitive consistency, a strong collaborative system and organizational consensus are required. The consistency of toolbars and operating attributes is very high. From the results, users think it should be so. People who have experienced tool design should have a deep feeling that systematic consistency is a basic and very systematic work. It requires highly abstract capabilities to define the consistency and implementation mechanisms of different modules and different products in order to achieve true consistency.

[0028] Scenario-based: As time goes by, the functions and help documents of poster design products tend to increase. Users tend not to look for them one by one. Due to the business attributes of some functions, it is difficult to make the product easy to use naturally in the short term. Therefore, it is necessary to consider how to convey the usage of the current functions to users and scenario-based the tutorials.

[0029] Operation visualization: Tool products have many operations and even many industry-specific expressions. So how to express some concepts clearly? For example, the parameters and annotations of poster elements cannot be expressed intuitively through text. Then you can quickly understand what they mean through graphics or animations. Assisting users to efficiently produce high-quality content is the goal of productivity tools. This goal will not change in the future. How to assist users to produce faster, simpler, and higher-quality content, so that the output content can play a greater value, is the trend of future tool development. Through artificial and intelligent AI Agents, this is an open question that deserves continuous exploration and thinking by every practitioner. Productivity tools are essentially content production tools. RAG will be able to directly produce content that was originally generated by people, reducing user interaction.

[0030] In the method for generating RAG posters in the embodiment of the present application, the form of input and output will change. It currently still exists in the form of a co-pilot, using RAG to assist the original product scene to assist the task flow of the main scene. In poster design, RAG does replace the original tool to produce content, dynamically collects information such as clicks and views of information posters, and combines it with business links involving upstream and downstream data to generate RAG poster design. Therefore, the form of RAG as a "co-pilot" is to allow the current tools to play a further value and make local operations more efficient and friendly. The method for generating RAG posters in the embodiment of the present invention provides a design tool for poster and material templates so that people who have design needs but no professional design capabilities can complete the design. Based on the design files, products, designs, and development can all be collaborative and downstream developed, and all project members can instantly obtain file data information.

[0031] In addition, no enterprise is willing to take the risk of data leakage and upload its own private domain data to a third-party platform for training. Therefore, the application solution that completely relies on the capabilities of the general large model itself is not feasible in the enterprise scenario. The private domain container in this application meets the requirements of data security.

[0032] The intelligent agent in the embodiment of the present application can automatically collect and analyze user feedback, and the private domain personalized model is similar to a search engine. It can solve the problem that the internal structure of the large model will be solidified at the moment when it is trained, so it is difficult to give an answer when asking the large model about some of the latest events.

[0033] In the embodiment of the present application, the learning framework provided is the structure of an information poster. A specific information poster generally consists of four parts: logo, title, main body, and background.

[0034] The quality of poster design will directly affect the user experience. In many cases, users will directly copy the poster layout of competitors to their own layout without considering the characteristics and ideas of their own posters. In theory, this approach will not make your own layout wrong. However, many people often overlook the starting point of others' design and whether it matches their own poster layout. If you don't understand these rashly copying designs, you will develop a bad habit over time, which will lead to insufficient stimulation of thinking and insufficient knowledge accumulation. When you need to carry out special design ideas, you will encounter many difficulties.

[0035] If you are just starting out in this field, you can refer to other people's designs in the early stages, but you must understand other people's design thinking, understand the reasons behind them, and accumulate them in your mind as your own knowledge reserves. We need to select appropriate information based on the direction of our own posters, as well as the user's behavior and characteristics, combined with the information environment, and organize and present it in a suitable way, so that users can obtain and understand the information more easily, and complete the organization and communication of information.

[0036] Related information needs to be classified. Regardless of the type of poster module, information layering should be done well in the design. When two content elements are related, they should be presented to users as a whole.

[0037] (1) The number of levels should be close to "3".

[0038] Information hierarchy is one of the important factors that affect the efficiency of information transmission on the page, so how to do a good job of information hierarchy on the page is particularly important. Generally, graphic design is designed around the four basic principles of contrast, alignment, intimacy, and repetition.

[0039] After an overall study of online layouts and concepts, we found that there is a simple rule based on the principles of typography design: the hierarchy of the main content is controlled at about three levels. A hierarchy of about three levels is the easiest for users to identify and is not easy to confuse visually. After three levels, the complexity increases proportionally with the number of levels. But in fact, we only need to classify the information and control the number of levels, and it will instantly become simpler and easier to understand. Therefore, before we start designing, we need to first think about a logic: try to control our content hierarchy to about three levels, and there must be a relatively obvious contrast relationship within these three levels.

[0040] The point is not to tell you such a simple truth, but when faced with complex hierarchies, how should we control its hierarchical performance so that it can be maintained at about three layers in the final presentation, allowing users to obtain information more efficiently.

[0041] (2) Source screening and processing.

[0042] The key to source screening (key selection) is that when taking over complex information, you first need to make a first-level judgment from the source to understand whether the information is really needed. The essence of source screening is to judge whether the information we are currently presenting is necessary. For example, sometimes the poster message gives us a lot of information, but we may not accept it all, but judge it based on the current summary scenario. Determine the rationality of each piece of information, so that you can decide whether some elements can be deleted.

[0043] (3) Reduce the complexity of the layout.

[0044] The essence of information arrangement is to control the complex hierarchy within a range of about three layers by subjectively organizing and reconstructing the information. This ensures the simplicity, regularity, and recognizability of our layout. Based on current practice, the methods of reasonably arranging information levels can be roughly divided into: grouping, component, organization, integration, and weakening (graphic design principles).

[0045] Information grouping is a form that everyone can think of when designing. Grouping can group complex information into a whole, thereby reducing the overall complexity and clarifying the clues. There are three main common grouping methods: spacing, dividing lines, and cards.

[0046] Using component splitting is an essential part of designing information layout, so the proficiency in this part of design is also the highest. These components play a relatively important role in the division of information levels: tree structure, table structure, step bar, and tab.

[0047] Flexible organization: By adjusting the organizational method, the multi-layer structure style can be controlled within three layers by splitting the layout.

[0048] Cleverly integrate, using the form of information integration to reduce its impact on the hierarchy. Information integration is essentially to integrate two levels of information into the same level through different combinations. For example, in the poster design, the title and the text are separated and arranged together, so that they become the same level, which effectively reduces the complexity of the layout. This can not only save visual space, but also reduce the complexity of the information level. However, in the process of information integration, we need to pay attention to: if placing them on the same level will cause misunderstandings, do not use this solution at this time.

[0049] Information weakening, the principle of information weakening is: weaken some parts that exceed the level. Therefore, we need to learn to classify information and weaken unimportant information, so that the overall presentation will be much better. But it is undeniable that there will still be more complex layout elements, even if they are reduced, there will still be too many levels, so at this time, we must remember to trace all these information from the source to see if they are necessary (graphic design principles).

[0050] (4) Highlight hot spots.

[0051] When displaying key elements, they need to be highlighted to help users quickly locate the target. When designing large-area text layouts, attention should be paid to the font thickness, which determines whether our design is easy to read. Whether it is the title or the content, fonts that are too heavy or too light will reduce the basic recognition of the text, and when there is too much text information, it is easy to cause visual fatigue if you focus on the recognition of the text for a long time. When there are multiple elements in the layout, these elements are prioritized to highlight the core functional elements. The user's action of browsing the layout is in the form of large-scale scanning, which means that we need to weaken irrelevant information to ensure the basic aesthetics of the layout and provide a good experience.

[0052] In actual applications, designers often face business demands raised by customers. When customers explain their demands and demand priorities, it is inevitable that they will be interfered by their subjective wishes, resulting in the fact that the functions actually implemented by the design cannot solve the customer's business problems, or new problems arise after the problems are solved. To solve the real problems of customers, it is necessary to deeply understand the customers and restore the business scenarios. Only by being able to simulate or participate in the business as a real customer can we find out the core cause of the problem. At the same time, based on the proportion of users affected by the problem in the same industry and the ability of competitors to solve the same problem, we can clearly identify the strategic value of the business to the platform, thereby giving priority to the demand and reasonably scheduling resources to solve it.

[0053] The technical solution of the present application is introduced below in conjunction with specific embodiments.

[0054] Please see attached Figure 1 , attached Figure 1 It is a flowchart of a poster generation method based on RAG poster design disclosed in an embodiment of the present application.

[0055] The poster generation method based on RAG poster design disclosed in this application may include: S1, taking the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, the RAG agent identifies the input poster visual design content element information and understands the semantics of the poster visual design content element information, uses the ReAct mode (the ReAct mode has the process of Thought, Action, Observation, and loop, and the role of the Executor is to execute this loop. The executor will always perform the event loop until the goal is solved or the number of thinking iterations exceeds the maximum number) as the starting point of the RAG agent design mode, and obtains the poster visual design generation background elements and automatic typesetting design parameters that match the poster visual design content element information; S2. Based on the poster visual design generating background elements and the automatic typesetting design parameters, displaying the target poster output by the poster design generating model in the second area of ​​the first interface; Among them, the RAG agent is obtained after training based on the personalized vector database configured in the private domain container as a training set and the basic LLM large model and basic literary graph model based on the public domain. The personalized vector database at least stores the promotional poster design knowledge template that conforms to the corporate culture content of the user. The RAG agent can at least reason and act to solve the language reasoning and decision-making tasks of diversified poster design in the personalized vector database.

[0056] The executing entity of the embodiment of the present application may be an intelligent AI Agent, especially a ReAct Agent. The ReActAgent may be based on the ReAct model as the starting point of the RAG intelligent agent design model, and by simulating the characteristics of human-computer "co-creation" in human thinking and action processes, it can automatically generate background elements for poster designs, such as background images.

[0057] In the embodiment of the present application, the poster visual design content element information does not require the user to make a professional expression as required by the large model output in the prior art. It only needs to input text information that needs to be reflected in the poster, and the intelligent agent can automatically reason and act to obtain a poster design generation background that matches it. In the application, you can search in the private domain container corresponding to the user to see if there is a corresponding poster visual design generation background element. If the most matching poster design generation background is not found in the private domain container, the intelligent agent's Wensheng graph model is called to generate the background. Among them, the input of the Wensheng graph model can be the information obtained by the intelligent agent after analysis.

[0058] The implementation method of the embodiment of the present application can be a RAG poster making tool. The RAG poster making tool can be software. The software can run in an operating system. The operating system can be a PC operating system or a mobile operating system, as long as it can run the AI ​​poster making tool in the embodiment of the present application.

[0059] In the application, a ReAct Agent needs to define the following elements.

[0060] Tools have two most important parameters, name and description.

[0061] Name is the function name, and description is the natural language description of the tool. LLM can decide whether to use the tool based on the description. The description of the tool should be very clear, explaining the function of the tool, when to use it, and when it is not applicable.

[0062] For example, we can simply concatenate two descriptions to provide the Agent with an arithmetic tool. The arithmetic tool function does not require actual code. The large model can complete the actual arithmetic operations based on its own reasoning ability. This is because they use natural language processing technology. Users can interact with these tools through natural language without writing any code. Of course, for more complex tool functions, detailed code construction is still required.

[0063] In the ReAct mode, the executor coordinates the components and guides the operation when the Agent is running. The ReAct mode has a process of Thought, Action, Observation, and a loop, and the executor's role is to execute this loop.

[0064] In an embodiment of the present application, after the RAG poster production tool is run, an interface may be displayed, and the interface may have a tab such as RAG poster design. When the user selects the tab, the corresponding design interface may pop up.

[0065] In the embodiment of the present application, the design interface may be the first interface. Figure 2 , attached Figure 2 1 is a schematic diagram of the display effect of the first interface shown in the embodiment of the present application. The first interface 10 in the embodiment of the present application may include a first area 101 and a second area 102. The content input in the first area 101 can interact with the poster design generation model. The second area 102 is a display area that can display the generation result of the poster design generation model.

[0066] In an embodiment of the present application, the poster visual design content element information may include the content that needs to be displayed on the poster. The user can enter a complete article in the first area 101, and the intelligent agent can automatically generate a corresponding poster design generation background based on the complete article. In the application, the user can also enter only brief content as the poster visual design content element information, for example, only enter simple descriptive words such as "retirement, yellow land, guardian, help farmers, main members", etc., and the embodiment of the present application can obtain the poster visual design content element information. In the application, the poster visual design content element information can be obtained, for example, after determining that the user input is completed or detecting that the user clicks the RAG description button 1011 in the first area. Determining that the user input is completed can be, for example, determining that there is no input action within a predetermined time.

[0067] In an embodiment of the present application, after obtaining the poster visual design content element information input by the user, such as the above brief content, the RAG agent identifies the input poster visual design content element information. As mentioned above, since LLM is not good at the field of posters, its output often does not conform to the professional content in this field. Therefore, in an embodiment of the present application, the RAG agent can be constructed based on the improvement of multiple basic models, and the ReAct mode can be used as the starting point of the RAG agent design mode. It can be based on the personalized vector database configured in the private domain container as a training set and based on the public domain basic LLM large model and the basic text map model for training.

[0068] In the embodiment of the present application, the private domain container can store each user's own personalized vector database, and the vector database stores vector data that can represent various templates, poster parameter data, etc.

[0069] In the application, the first area may have a prompt word, such as "can be manually fine-tuned to obtain better results", to prompt the user to manually adjust the poster generation information, so as to control the generated result to better meet the user's own needs.

[0070] In the embodiment of the present application, the first interface may further include a third area 103 and a parameter setting area, wherein the parameter setting area is used to set certain parameters for generating a poster, for example, the image size of the generated poster, the number of generated posters, and the like.

[0071] In the embodiment of the present application, the first interface may also be provided with a generate button, and by clicking the generate button, a target poster can be generated and displayed in the second area.

[0072] In the embodiment of the present application, the poster design generation model is obtained after training based on the personalized database configured in the private domain container as a training set and the basic LLM large model and basic text graph model based on the public domain. It can output professional prompt words in the field of poster design according to natural language, and because the private domain container can ensure that the enterprise's data is not uploaded to the public network, it can ensure that the privacy of enterprise users is not leaked.

[0073] In the application, the personality vector database at least stores the promotional poster knowledge template that conforms to the user's corporate culture content. As mentioned above, the personality vector database can be a distributed database, and a large number of information poster templates can be stored in the database, so as to serve as a training set and a plug-in for a large model and a basis for controlling the text graph model. Among them, the promotional poster knowledge template can be a poster template of the promotional idea that conforms to the user's corporate culture content, and the data is internalized into the model parameters through the industry information poster feature data and knowledge and the industry expert promotion knowledge. Therefore, it can guide the poster design generation model to output a poster design generation background that meets the user's needs.

[0074] It is understandable that in the embodiment of the present application, the automatic typesetting design parameters may include various necessary parameters for designing a poster, such as the position of each paragraph of text in the poster, the font size, font, and color. It is understandable that a poster is not a picture, and posters require relevant knowledge of graphic design. Therefore, the generated background should also conform to the concept of graphic design.

[0075] The specific method for generating the optimal layout design area of ​​the poster is: constructing an optimal layout design area model, and determining the optimal layout design area according to the optimal layout design area model.

[0076] Furthermore, the expression of the optimal area model Obj for typesetting design is:

[0077] Wherein, w_li represents the length of the layout design rectangle, h_li represents the width of the layout design rectangle, and L represents the number of layout design rectangles.

[0078] Furthermore, in the step, the aspect ratio r_li of the optimal layout design area is calculated. If r_li>1, the poster is laid out in horizontal layout, otherwise it is laid out in vertical layout.

[0079] The embodiment of the present application can generate a target poster that meets the user's true intention while ensuring that the user's privacy is not leaked and the user's operation convenience is ensured. Since the software is deployed in the operating system, the user's personalized needs can be flexibly adjusted.

[0080] In the embodiment of the present application, a template can also be selected for reference of the poster design generation model. The software tool can display a template display interface, which can include multiple templates for selection, and multiple templates can be sorted by real-time access data. Since the posters generated based on the templates are likely to be accessed in real time, the embodiment of the present application can be sorted according to the real-time access data of the templates, and the real-time access data can be obtained from the log data. The template can be a template corresponding to the poster previously generated by the intelligent agent, and the template can have a background and the poster content edited in the background.

[0081] In the embodiment of the present application, multiple category elements can be set, such as: new media, red propaganda, news information, good morning, long pictures, etc., and multiple sorted templates can be displayed under each category. The sorting rule of the template is the above-mentioned real-time access data corresponding to each template. The access data can be the number of citations, clicks, visits, etc., which can be automatically set according to actual needs.

[0082] The user can select a target template in the template display interface. When it is determined that the user has selected the target template, the current display interface jumps to the first interface, and the target template is displayed in the second area of ​​the first interface. Of course, the title and content corresponding to the target template can also be displayed in the first area of ​​the first interface. When the user rewrites the content in the first area, a target poster that conforms to the style of the target template can be further generated.

[0083] Please continue to see Figure 3 , Figure 3 This is another schematic diagram of the display effect of the first interface shown in the embodiment of the present application.

[0084] In an embodiment of the present application, the third area 103 of the first interface is used to add a reference poster. The reference poster can make the poster more predictable based on past experience, and can control the output of the RAG agent, which can be based on the various layout parameters of the reference poster, such as the theme, logo, text, and background, etc., which are used to make the poster as a reference, and generate a target poster based on the aforementioned generated poster generation information. Due to the addition of the reference poster, compared with a similar poster generated without adding the reference poster, that is, the target poster, the style of the target poster can be closer to the style of the added poster, but it is not completely consistent, but adding control factors to the layout parameters of the RAG agent can be more in line with user expectations.

[0085] In the application, the user can click on the add button set in the third area, for example, a plus button is set in the third area. After clicking the plus button, the user can manually upload a reference poster or select a poster template displayed in the template library, or click the random generate button corresponding to the third area to randomly generate a poster.

[0086] Please continue to see Figure 4 , Figure 4 It is a schematic diagram of the display effect of the modification menu in the second area of ​​the first interface in an embodiment of the present invention.

[0087] like Figure 4 As shown, Figure 4 The second area 102 in the figure displays the target poster generated by the RAG agent or the target poster corresponding to the target template selected by the user.

[0088] In the embodiment of the present application, the poster elements of the poster that has been generated can be modified by the user. The poster elements may include a logo, title, subject and / or background. The user can modify the poster elements as needed. For example, clicking on the target poster displayed in the second area can pop up a pop-up window such as Figure 4 The modification menu shown can display various modification options that can be operated, such as RAG rewriting or modification options for adjusting the display mode. RAG rewriting can be, for example, uploading locally, sketching, continuing, rewriting, overwriting, expanding, polishing, deleting, etc., and adjusting the display mode can be, for example, creating a copy, undoing, pinning to the top, pinning to the bottom, locking, unlocking, transparency, clearing the canvas, etc.

[0089] In the application, users can modify the generated target poster. For example, to modify a certain text in the content, they can first select the text to be selected, and then select RAG Rewrite in the pop-up modification menu, and the RAG agent can rewrite the text. When Lock is selected in the modification menu, the selected content can be locked.

[0090] According to the above, the target poster can be divided into multiple levels to display the content. In the embodiment of the present application, the level of the generated target poster can be a preset number of levels, and the preset number of levels can be 3, for example, wherein each preset level can correspond to a poster layout, and each poster layout needs to have an associated element relationship that meets the comparison conditions. Comparison conditions such as comparison, alignment, intimacy, repetition and other comparison conditions can determine the comparison condition parameters of each page according to the generation order of each page in the target poster. The comparison condition parameters can be the aforementioned poster generation parameters. If the comparison condition parameters of the latter page meet the preset values ​​in the comparison conditions, or the values ​​obtained based on the calculation results of the preset function between the two comparison condition parameters meet the range. It is only necessary to be able to obtain whether the comparison conditions are met between the poster layouts through the comparison conditions. This can make the pages have a strong correlation, and due to the limitation of the levels, the poster is easier to catch the reader.

[0091] In some embodiments, the first area of ​​the first interface of the first interface also includes a clearing word input component for inputting clearing words, in which the user can input content that needs to be avoided, so as to prevent the RAG agent from generating content that matches the input clearing word. For example, if the clearing word includes a woolly bear, then the generated poster generation information will not include any content related to a woolly bear, and the generated target poster will not include any images related to a woolly bear.

[0092] In some embodiments, the embodiments of the present application can also perform polishing operations on the poster generation information that has been generated. In order to attract the user's attention, a very intuitive design displays the core features, a unified layout and structure, reduces the reader's learning and understanding costs, helps eliminate the reader's doubts, and makes these information elements clear and orderly displayed in front of the reader and improves the efficiency of use. For example, the user selects the content that needs to be adjusted in the poster generation information, and then selects a polishing operation in the pop-up menu, such as a rewrite operation or a manual modification operation. The rewrite operation can be to call the RAG intelligent body to rewrite and modify, and the manual operation can be manually rewritten according to the actual needs of the user.

[0093] It should be noted that in the embodiments of the present application, the LLM large model and the Vincent graph model can be integrated into the RAG intelligent agent, and a poster can be generated through the Vincent graph model based on the output of the LLM large model. Unlike generating pictures, generating pictures only needs to consider the background and image elements, while posters not only need to consider the layout, but also the layout between each poster element and the relationship between each poster layout. Therefore, the method of only using Vincent graphs to generate pictures and then rewriting them into posters cannot serve as the basis for the poster generation method in the present application.

[0094] In the embodiment of the present application, a background image can also be uploaded as the background image for making a poster, and a poster can be generated based on the uploaded background image.

[0095] In addition, in order to facilitate user operations, the setting method of each component of the software tool in the embodiment of the present application follows the principle of functional proximity, that is, a function used by the user corresponds to the component of the function corresponding to the next operation he needs to perform, for example, related functional modules or components are placed together so that they can interact and maintain more easily.

[0096] On the other hand, an embodiment of the present application also discloses a method for generating a RAG poster.

[0097] See also Figure 5 , Figure 5 It is another flowchart diagram of a method for generating a RAG poster shown in an embodiment of the present application.

[0098] The method for generating a RAG poster disclosed in this application includes: S500, taking the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, the RAG agent identifies the input poster visual design content element information and understands the semantics of the poster visual design content element information, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains the poster visual design generation background element matching the poster visual design content element information; S501, receiving a manual design operation performed by the user using the poster visual design content element information within the poster visual design generation background element to obtain a target poster.

[0099] Different from the aforementioned automatically generated posters, in the embodiment of the present application, the RAG poster making tool can also meet the manual design requirements of professionals.

[0100] In the application, the user can also input the poster visual design content element information based on the RAG poster making tool in the embodiment of the present application. The RAG agent can be the same as the AI ​​agent in the aforementioned embodiment, and can also automatically generate a poster design background.

[0101] Of course, when the user needs to design manually, he can also select a background image first. The background image can be uploaded by the user in the background upload component in the first interface, or the uploaded background image can be directly used. The background image is displayed in the second area of ​​the first interface overlaying the automatically generated poster design generation background.

[0102] When the user manually designs a poster, the user can click on the text, image and other component buttons corresponding to the second area to implement the poster design in the second area, such as adding a title, weather forecast component, content, etc. When selecting a background image, parameters can be set, and the parameter setting can be a setting for the image size. For example, mobile phones, tablets, computers, etc. can directly correspond to different platforms and set different image sizes, so that the background image is processed and displayed in the second area in a manner that conforms to the parameter setting.

[0103] The embodiment of the present application can also modify the style of the poster elements in the background image, and the user can select the style that can be presented according to the user's design goal. Of course, the canvas size can also be adjusted.

[0104] In the application, after the background is generated, it can be automatically adjusted to the first interface. The user can enter the description information and the RAG agent will generate the corresponding poster generation information.

[0105] It can be understood that other components of manual editing and the editing form of the content can refer to the aforementioned embodiments and will not be described in detail here.

[0106] It can be seen that in the embodiment of the present application, the software tool provided can avoid the interaction barriers caused by the web version and generate a target poster that meets the user's expectations while improving the user's operation convenience, and there are related element relationships between the various poster layouts of the target poster.

[0107] Corresponding to the aforementioned method embodiment, the present application also discloses a device for generating a RAG poster.

[0108] See also Figure 6 , Figure 6 It is a structural schematic diagram of a RAG poster generation device disclosed in an embodiment of the present application.

[0109] A RAG poster generation device disclosed in an embodiment of the present application includes: The acquisition module 61 is used to obtain the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, the RAG agent recognizes the input poster visual design content element information and understands the semantics of the poster design content element, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains the poster design generation background and automatic typesetting design parameters that match the poster design content element information; A generation module 62 is used to display the target poster output by the poster design generation model in the second area of ​​the first interface based on the poster design generation background and the automatic typesetting design parameters; wherein the RAG agent is obtained after training based on the personalized vector database configured in the private domain container as a training set and the basic LLM large model and basic literary graph model based on the public domain, and the personalized vector database at least stores a promotional poster design knowledge template that conforms to the corporate culture content of the user, and the RAG agent can at least reason and act to solve the language reasoning and decision-making tasks of diversified poster design based on the personalized vector database.

[0110] In some embodiments, before obtaining the description information, the device further includes a template processing module, which is used to: Display the current display interface as a template display interface; Receive a template selection operation of the user in the template display interface; wherein the data source element of the display template displayed in the template display interface is the personalized database, each of the display templates corresponds to at least one category element, and each display template under each category can be sorted based on the real-time access data of each template; After determining the target template based on the template selection operation, the current display interface is updated to display the first interface.

[0111] In some embodiments, after obtaining the poster generation information matching the description information, the device is further used to: Acquire a reference poster added by the user in the third area of ​​the first interface; Then, based on the poster generation information, the target poster output by the RAG agent is displayed in the second area of ​​the first interface, including: sending the poster generation information and the reference poster to the RAG agent, so that the RAG agent controls the typesetting parameters of the generated result based on the reference poster; the typesetting parameters are used for typesetting the theme, logo, text and background in the poster.

[0112] In some embodiments, the apparatus further comprises a modification module, configured to: Receiving a modification operation of a poster element in the target poster by a user; the poster element includes a logo, a title, a main body and / or a background; Determine the modification mode of the poster element by the modification operation; Modify the poster element based on the modification method; the modification method includes calling the RAG agent to rewrite the text or adjust the display method of the target element.

[0113] In some embodiments, determining the modification mode of the poster element by the modification operation includes: Receive a user's selection operation on the poster element; A modification menu pops up at the location of the selection operation; the modification menu includes modification options for rewriting the RAG or adjusting the display mode; Receiving a selection operation of the user on a modification option in the modification menu; A modification method corresponding to the poster element is determined based on the modification option corresponding to the selection operation.

[0114] In some embodiments, the target poster includes a preset number of layers; wherein each of the preset number of layers corresponds to a poster layout, and each poster layout has at least an associated element relationship that meets the comparison conditions.

[0115] In some embodiments, a clearing word input component for inputting clearing words is also provided in the first area of ​​the first interface, and the clearing words in the clearing word input component are used to control the content that cannot appear in the poster generation information output by the RAG agent.

[0116] In some embodiments, a polishing module is further included for: Receiving a retouching operation on the poster generation information; According to the polishing operation, the poster generation information is modified; the polishing operation includes a rewriting operation or a manual modification operation of selecting part of the information of the poster generation information.

[0117] The implementation process of each module in the device in the present application can refer to the implementation methods of each step in the aforementioned method embodiment, which will not be described in detail here.

[0118] The present application also provides an electronic device, see Figure 7 , Figure 7 The electronic device 1000 includes a memory 1010 and a processor 1020 .

[0119] The processor 1020 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.

[0120] The memory 1010 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. Among them, ROM can store static data or instructions required by the processor 1020 or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at runtime. In addition, the memory 1010 may include any combination of computer-readable storage media, including various types of semiconductor storage chips (such as DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory 1010 may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (such as a DVD-ROM, a double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (such as an SD card, a mini SD card, a Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0121] The memory 1010 stores executable codes, and when the executable codes are processed by the processor 1020 , the processor 1020 can execute part or all of the methods described above.

[0122] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.

[0123] Alternatively, the present application can also be implemented as a computer-readable storage medium (or non-transitory machine-readable storage medium or machine-readable storage medium) on which executable code (or computer program or computer instruction code) is stored. When the executable code (or computer program or computer instruction code) is executed by a processor of an electronic device (or server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.

[0124] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A poster generation method based on RAG poster design, characterized in that: include: The RAG agent uses the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, recognizes the input poster visual design content element information and understands the semantics of the poster visual design content element information, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains the poster visual design generation background elements and automatic typesetting design parameters that match the poster visual design content element information; Based on the poster visual design generating background elements and the automatic typesetting design parameters, displaying the target poster output by the poster design generating model in the second area of ​​the first interface; Among them, the RAG agent is obtained after training based on the personalized vector database configured in the private domain container as a training set and the basic LLM large model and basic literary graph model based on the public domain. The personalized vector database at least stores the promotional poster design knowledge template that conforms to the corporate culture content of the user. The RAG agent can at least reason and act to solve the language reasoning and decision-making tasks of diversified poster design.

2. The method according to claim 1, characterized in that Before obtaining the poster visual design content element information, the method further includes: Display the current display interface as the template design display interface; Receive a template selection operation of the user in the template display interface; wherein the data source element of the display design template displayed in the template display interface is the personalized vector database, each of the display templates corresponds to at least one category element, and each display design template under each category can be sorted based on the real-time access data of each template; After determining the target template based on the template selection operation, the current display interface is updated to an automatically designed and generated interface.

3. The method according to claim 1, characterized in that After obtaining the poster visual design content element information, the method further includes: Acquire a reference poster added by the user in the third area of ​​the first interface; Then, based on the poster generation information, the target poster output by the RAG agent is displayed in the second area of ​​the first interface, including: sending the poster generation information and the reference poster to the RAG agent, so that the RAG agent controls the typesetting parameters of the generated result based on the reference poster; the typesetting parameters are used for typeset design of the theme, logo, text and background elements in the poster.

4. The method according to claim 1, characterized in that: The method further comprises: Receiving a modification operation of a certain poster element in the target poster by a user; the poster element includes a logo, a title, a main body and / or a background element; Determine the modification mode of the poster element by the modification operation; Modify the poster element based on the modification method; the modification method includes calling the RAG agent to rewrite the text or adjust the display method of the target element.

5. The method according to claim 4, characterized in that The determining a modification mode of the poster element by the modification operation includes: Receive a user's selection operation on the poster element; A modification menu pops up at the location of the selection operation; the modification menu includes modification options for rewriting the RAG or adjusting the display mode; Receiving a selection operation of the user on a modification option in the modification menu; A modification method corresponding to the poster element is determined based on the modification option corresponding to the selection operation.

6. The method according to any one of claims 1 to 5, characterized in that: The target poster includes a preset number of layers; wherein each of the preset number of layers corresponds to a poster layout, and each poster layout has at least an associated element relationship that meets the comparison conditions.

7. The method according to any one of claims 1 to 5, characterized in that: A clearing word input element component for inputting clearing words is also provided in the first area of ​​the first interface. The clearing words in the clearing word input component are used to control the element contents that cannot appear in the poster generation information output by the RAG agent.

8. The method according to any one of claims 1 to 5, characterized in that: Also includes: Receiving a retouching operation on the poster generation information; Modifying the poster generation information according to the polishing operation; The polishing operation includes a rewriting operation or a manual modification operation of selecting part of the poster generation information.

9. The poster generation method based on RAG poster design is characterized by: include: The RAG agent uses the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, recognizes the input poster visual design content element information and understands the semantics of the poster visual design content element information, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains the poster visual design generation background element that matches the poster design content element information; A manual design operation performed by the user using the poster visual design content element information within the poster visual design generation background element is received to obtain a target poster.

10. A poster generation device based on RAG poster design, characterized in that: include: An acquisition module is used to obtain the poster visual design content element information input by the user in the first area of ​​the first interface of the RAG poster making tool as a trigger, the RAG agent recognizes the input poster visual design content element information and understands the semantics of the poster design content element, uses the ReAct mode as the starting point of the RAG agent design mode, and obtains the poster design generation background and automatic typesetting design parameters that match the poster design content element information; A generation module displays the target poster output by the poster design generation model in the second area of ​​the first interface based on the poster design generation background and the automatic typesetting design parameters; wherein the RAG agent is obtained after training based on the personalized vector database configured in the private domain container as a training set and the basic LLM large model and basic literary graph model based on the public domain, and the personalized vector database at least stores a promotional poster design knowledge template that conforms to the corporate culture content of the user, and the RAG agent can at least reason and act to solve the language reasoning and decision-making tasks of diversified poster design based on the personalized vector database.

11. An electronic device, comprising: at least one processor; And, a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a method for generating a RAG poster as described in any one of claims 1-8 or a method for generating a RAG poster as described in claim 9.

Images