System that calculates music source popularity potential and provides real-time arrangement optimization guidance through AI-based multimodal data fusion

Abstract

The present invention relates to an artificial intelligence-based music production support system, and more specifically, to a system and method that multimodally fuses a plurality of heterogeneous data including audio signals, lyric text, and short-form platform response data to calculate a viral potential index and an audio-vocal compatibility index for a sound source, and provides a timeline arrangement improvement guide in real time by linking with a digital audio workstation.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence-based music production support system, and more specifically, to a system and method that multimodally fuses a plurality of heterogeneous data including audio signals, lyric text, and short-form platform response data to calculate a viral potential index and an audio-vocal compatibility index for a sound source, and provides a timeline arrangement improvement guide in real time by linking with a digital audio workstation. Background Technology

[0002] In the conventional music planning and production environment, the potential for a song's success and the direction of the arrangement were determined based on the subjective hearing and experience of the workers. Recently, there have been some attempts to predict success by extracting simple audio characteristics such as tempo or energy, but this has limitations in that the accuracy of predicting success in the actual short-form environment is significantly lower because it fails to comprehensively reflect the song's timeline development structure (e.g., length of the intro) or the rapidly changing trend momentum unique to social media platforms.

[0003] Furthermore, while general spectrum analyzers are used during the mixing and mastering stages to check for frequency conflicts between audio sources and vocal sources, they merely provide visual information regarding overlapping frequency bands. Since the degradation of auditory clarity caused by excessive compression (the use of compression and limiters), which frequently occurs in modern commercial music, cannot be quantified as objective physical figures, non-professionals often lose the balance between sound pressure and vocal clarity when working on the project.

[0004] Furthermore, when attempting to visualize and provide such complex analysis data to users in real-time within a digital audio workstation environment, the screen attempts to refresh with every minute parameter manipulation by the operator. This triggers resource contention between the audio signal processing thread and the graphics rendering thread, causing spikes in system usage and consequently leading to audio dropouts. Consequently, there existed a critical technical barrier that hindered an uninterrupted music production environment. The problem to be solved

[0005] The main problem that the present invention aims to solve is to overcome the limitations of subjective evaluation and simple audio analysis of conventional technology and to provide a system that derives the potential for viral success of a sound source as a quantitative index by fusing a viewing retention rate algorithm of a short-form platform with changes in public trends through mathematical modeling.

[0006] Another problem that the present invention aims to solve is to provide a precise audio source-vocal compatibility index that conforms to global audio distribution standards by analyzing the dynamic range and loudness ratio of an audio signal and determining the degradation of clarity caused by excessive compression as a physical value.

[0007] Another problem that the present invention aims to solve is to prevent system overload and provide a stable real-time arrangement improvement guide by dynamically controlling whether the screen is updated based on user parameter manipulation through a threshold value derived based on genre-specific statistical data.

[0008] The problems of the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0010] According to an embodiment of the present invention for solving the above problem

[0011] An audio source input unit that receives an audio signal subject to multimodal analysis from a digital audio workstation or an external storage medium;

[0012] A trend weight learning module that calculates dynamic trend momentum weights of the market by combining response data from social media and short-form platforms;

[0013] A key source feature extraction module that derives key source features, including an initial dynamic change amount and a hook section start time, from the received audio signal and metadata;

[0014] A lyric feature extraction module that derives lyric features including the number of syllables, the ratio of repeated phrases, and the ratio of plosive series occurrences from lyric text corresponding to the above audio signal;

[0015] A feature fusion module that generates fusion features by normalizing and combining the above-mentioned main source features and the above-mentioned lyrics features;

[0016] A potential for success index calculation module that calculates a viral potential index of a song (a form of potential for success index) using the dynamic trend momentum weights and the fusion features;

[0017] A sound source-vocal compatibility index analysis module that derives a sound source-vocal compatibility index by quantifying the degradation of clarity caused by physical frequency conflicts between vocal and accompaniment sources and excessive sound pressure;

[0018] A real-time parameter change tracking module that determines whether the screen refresh engine is operating by continuously monitoring the amount of change in the operator's parameter manipulation;

[0019] A production guide generation module that generates a guide card including timeline adjustment and sound adjustment guides based on the fluctuation results of the indices calculated above; and

[0020] An information visualization output unit that visually outputs the guide card and index display information according to the screen refresh permission signal of the above parameter change real-time tracking module;

[0021] Includes,

[0022] The above trend weight learning module is,

[0023] Collects behavioral indicators including the usage of challenge templates for specific songs, the elapsed time of data collection, and data on market competition intensity during the same period from an external platform, and applies an exponential decay logic over time based on the said elapsed time of collection,

[0024] Determining whether the ratio of the above behavioral indicators to the above collected market competition intensity exceeds a preset abuse detection threshold, and

[0025] If the above calculation ratio is determined to exceed the above abuse detection threshold, the above calculation ratio is characterized by applying a logarithmic function-based correction operation to forcibly suppress the increase of the above dynamic trend momentum weight, thereby blocking metric distortion caused by artificial traffic overheating of a specific fandom group.

[0026] The above abuse detection threshold is set to a constant value of 10.0, and

[0027] The above trend weight learning module drives the calculation logic by inputting the moving average value of the last 30 days as a substitute competition intensity to prevent system errors when an exceptional situation is detected in which the reception of the above market competition intensity data fails, and

[0028] It is characterized by preemptively preventing zero-division operation errors by forcibly fixing the dynamic trend momentum weight calculated as a result of applying the above exponential decay logic to a lower limit value of 0.1 when the result is less than 0.1.

[0029] The above-mentioned main source feature extraction module derives the initial dynamic change amount by calculating the square root mean square level change range at the initial 10-second point relative to the start point of the audio signal waveform in decibels, and determines the start point of the hook section where the spectral energy forms a local maximum value in absolute time units.

[0030] The above-mentioned box office potential index calculation module is,

[0031] A compensation value proportional to the initial dynamic change amount is added to the value obtained by multiplying the similarity between the feature vector of the input sound source and the trend reference vector by the dynamic trend momentum weight, and

[0032] It determines whether the start time of the above-determined hook section exceeds 15 seconds, which is a dropout threshold set based on the average viewer dropout rate data of the short-form platform, and

[0033] If the above determination result exceeds 15 seconds, the viral potential index is recalculated by applying an inverse decay constant proportional to the time exceeding 15 seconds, thereby inducing changes to the timeline arrangement structure, such as shortening the intro, to increase the defense rate of the initial viewing retention rate algorithm.

[0034] The above-mentioned box office potential index calculation module is,

[0035] Additionally collect creator diffusion metrics including the number of user-generated content generated from the audio signal from external social platforms, and continuously monitor whether the collected creator diffusion metrics exceed a preset saturation threshold,

[0036] When the monitoring results determine that the saturation threshold is exceeded, the method is characterized by mathematically modeling and reflecting the phenomenon of accumulated public meme fatigue, where the algorithm forcibly applies a diversity quota to lower exposure frequency when a specific sound source monopolizes the platform feed, by calculating the increase curve of the viral potential index by activating an exponential saturation suppression logic so that the increase curve converges to 1.0.

[0037] The above-mentioned box office potential index calculation module is,

[0038] In order to internally identify an exceptional situation in which the hook section, which is a distinct energy surge point, is not detected because the above audio signal corresponds to an ambient genre or a classical genre, the spectral center variance flatness of the audio signal is calculated, and

[0039] If the calculated flatness satisfies the preset short-form evaluation exclusion criteria, the logic for applying an inverse decay constant due to exceeding 15 seconds at the start of the hook section is immediately bypassed and the system automatically switches to a pure streaming evaluation mode, thereby blocking the penalty for reducing the viral potential index due to the absence of the hook section and defending against calculation errors by adding a constant to the denominator of the time variable. Effects of the invention

[0041] The system of the present invention for calculating the potential for success of music and providing a real-time arrangement improvement guide through AI-based multimodal data fusion to solve the above-mentioned problems provides the following effects.

[0042] By operating the exponential decay logic over time and the logarithmic correction calculation for the ratio relative to market competition intensity, it is possible to block indicator distortion caused by streaming offensives from specific fandoms and secure highly reliable dynamic trend momentum weights.

[0043] If the point of entry into the hook section exceeds the platform's exit threshold, an inverse decay constant is applied to downgrade the potential index and guidelines such as shortening the intro are provided, thereby having the effect of inducing the worker to improve the song into a timeline structure that meets the platform's viewing retention rate standards.

[0044] When frequency masking occurs, the loudness and dynamic range ratio of the corresponding section is determined and exponential attenuation correction is strongly applied, thereby warning of the perceived muffled sound that occurs when an operator excessively raises the sound pressure through an intuitive numerical drop, and thus helping to achieve clear sound quality that meets commercial standards.

[0045] By performing closed-loop control that immediately blocks screen updates and switches to internal accumulation mode when the amount of parameter manipulation change is below the minimum significant change threshold, it has the effect of dramatically reducing rendering load and providing a pleasant real-time music production environment without audio interruptions.

[0046] The effects according to the present invention are not limited to those exemplified above, and a wider variety of effects are included within the present invention. Brief explanation of the drawing

[0048] Figure 1 illustrates the overall flowchart of the AI-based music success potential calculation and guide system according to the present invention. Figure 2 illustrates a flowchart for calculating dynamic trend momentum according to the present invention. Figure 3 illustrates a flowchart of a major source-based viral potential evaluation according to the present invention. Figure 4 illustrates a flowchart of the audio source-vocal compatibility index (MVCI) analysis according to the present invention. Figure 5 illustrates a flowchart of real-time guide efficiency and UI load control according to the present invention. FIG. 6 is an example of a user interface that parses a project track through the audio source input unit (110) and performs timeline analysis and arrangement control based on virtual audio random access technology. FIG. 7 is an example of a virtual parameter manipulation interface placed at the bottom and a real data-based viral possibility trajectory curve derived from the above-mentioned box office potential index calculation module (140). FIG. 8 is an example screen diagram showing the parameter change real-time tracking module (160) detecting the minimum significant change amount (ART standard) and quantifying and outputting the final viral potential index. FIG. 9 is an example of a source input screen in which the above-mentioned sound source-vocal compatibility index analysis module (150) receives vocal stem and accompaniment stem files and starts a precise diagnosis. FIG. 10 is an example graph showing that the sound source-vocal compatibility index analysis module (150) analyzes the masking phenomenon by frequency band and calculates the maximum masking frequency and the final compatibility index. FIG. 11 is an example of an integrated producing improvement guide prescription generated by the modification guide generation module (170) for lyrics, structure, mixing, and distribution strategies. FIG. 12 is an example screen diagram showing the information visualization output unit (180) outputting in detail audio short form indicators, lyrics hooking index, initial skip defense power, and editorial playlist entry probability. Figure 13 is an example of a result visualized in the form of a gauge chart of the final comprehensive viral potential index derived by fusing sound, text, social, and market data by the system. Specific details for implementing the invention

[0049] Hereinafter, various embodiments are described in more detail with reference to the attached drawings. The embodiments described in this specification may be modified in various ways. Specific embodiments may be depicted in the drawings and described in detail in the detailed description. However, specific embodiments disclosed in the attached drawings are intended only to facilitate understanding of various embodiments. Accordingly, the technical concept is not limited by specific embodiments disclosed in the attached drawings, and it should be understood that it includes all equivalents or substitutions that fall within the spirit and scope of the invention.

[0050] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another.

[0051] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. When a component is described as being “connected” or “connected” to another component, it should be understood that it may be directly connected to or connected to that other component, or that there may be other components in between. On the other hand, when a component is described as being “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

[0052] Meanwhile, a "module" or "part" for a component as used in this specification performs at least one function or operation. Furthermore, a "module" or "part" may perform a function or operation by hardware, software, or a combination of hardware and software. Additionally, a plurality of "modules" or a plurality of "parts," excluding a "module" or "part" that must be performed on specific hardware or on at least one processor, may be integrated into at least one module. A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0053] In addition, power, power transmission, and control therefor for the following assembly configurations and embodiments, including "by control," follow conventional technology including terminals, applications, hardware control modules, etc., so they are omitted to avoid redundancy.

[0054] In addition, the operation embodiments and configurations described in a general manner without being explained in detail below follow the prior art and are omitted in order to focus on describing the purpose of the present invention and the resulting effects.

[0055] Furthermore, in describing the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description is abbreviated or omitted.

[0056] The system for calculating the potential for success of a song and providing a real-time arrangement improvement guide through artificial intelligence-based multimodal data fusion according to the present invention may include an audio source input unit (110), a trend weight learning module (120), a main source feature extraction module (130), a lyric feature extraction module (135), a feature fusion module (137), a success potential index calculation module (140), a song-vocal compatibility index analysis module (150), a parameter change real-time tracking module (160), a production guide generation module (170), and an information visualization output unit (180).

[0057] The above audio source input unit (110) receives an audio signal to be analyzed from a digital audio workstation (hereinafter referred to as “”) or an external storage medium.

[0058] The above trend weight learning module (120) calculates dynamic trend momentum weights by combining reaction data from social media and short-form platforms.

[0059] The above-mentioned main source feature extraction module (130) derives main features including the initial dynamic change amount and the hook section start time from the audio signal and metadata.

[0060] The feature fusion module (137) generates a fusion feature by combining the main source feature and the lyrics feature. The viral potential index calculation module (140) calculates a viral potential index using the dynamic trend momentum weight and the fusion feature.

[0061] The above-mentioned sound source-vocal compatibility index analysis module (150) calculates a sound source-vocal compatibility index by quantifying the frequency overlap and over-compression state between the vocal source and the accompaniment source.

[0062] The above parameter change real-time tracking module (160) monitors the amount of parameter manipulation change and screen load indicators to determine whether to allow screen updates.

[0063] Additionally, in this specification, the “screen load indicator” may be defined as an indicator provided by an operating system or execution environment, including (i) the degree of frame throughput reduction of a graphics rendering thread and (ii) system utilization, and the screen refresh cycle may be configured to be extended when the screen load indicator exceeds a reference value.

[0064] The above production guide generation module (170) generates a guide card including a timeline adjustment and sound adjustment guide based on the fluctuations of the viral potential index and the sound source-vocal compatibility index.

[0065] The information visualization output unit (180) visually outputs the guide card and index display information according to the screen update allow signal.

[0066] According to one embodiment, the lyrics feature extraction module (135) receives lyrics text corresponding to an input sound source, normalizes symbols and spaces in the lyrics text, and then calculates lyrics features including (i) the total number of syllables, (ii) the repetition ratio (the ratio of phrases in which the same phrase is repeated two or more times among all phrases), and (iii) the occurrence ratio of plosive series letters (the occurrence ratio of plosive series letters among all letters). The lyrics features are normalized so that each item is in the range of 0 to 1, and are transmitted to the feature fusion module (137) as a set of normalized values.

[0067] According to one embodiment, the feature fusion module (137) normalizes the main source feature derived from the main source feature extraction module (130) and the lyrics feature derived from the lyrics feature extraction module (135) to a range of 0 to 1, respectively, and then sequentially connects the normalized features to generate fusion features. The fusion features are transmitted to the success potential index calculation module (140) and combined with dynamic trend momentum weights to be used for calculating the viral potential index.

[0068] More specifically, the audio source input unit (110) receives audio sample buffer and timeline information of a master bus or submix bus within a project through a plugin extension protocol of a digital audio workstation (hereinafter referred to as “”). The audio source input unit (110) buffers the entire section of the project by dividing it into multiple time windows regardless of the audio playback position; in one embodiment, the time windows may be set to a length corresponding to 1024 samples, and the time windows may be shifted by an overlap ratio of 50%. The audio source input unit (110) normalizes and divides the audio data and transmits it to the main source feature extraction module (130).

[0069] Additionally, the trend weight learning module (120) collects behavioral indicators including challenge template usage, collection elapsed days, and market competition intensity through an application interface of an external platform, and calculates dynamic trend momentum weights by applying a log correction operation to the damping coefficient according to the collection elapsed days and the calculation ratio of market competition intensity relative to the behavioral indicators. In one embodiment, if the collection elapsed days are 3 days and the calculation ratio exceeds 10.0, the trend weight learning module (120) may apply the log correction operation to limit the fluctuation range of the dynamic trend momentum weights.

[0070] The above-mentioned main source feature extraction module (130) performs frequency analysis on the audio waveform received from the audio source input unit (110) and calculates the initial dynamic change amount and the hook section start time. The initial dynamic change amount is defined as the root mean square level change range at the initial 10-second point relative to the start point of the audio waveform and is calculated in decibels. The hook section start time is calculated by determining the point in time when the spectral energy forms a local maximum value as an absolute time in seconds.

[0071] The above-mentioned success potential index calculation module (140) calculates a viral potential index using the fusion feature quantity generated by the above-mentioned feature fusion module (137) and the dynamic trend momentum weight calculated by the above-mentioned trend weight learning module (120). If the hook section start time exceeds the breakout threshold (15 seconds), a damping operation linked to the excess time is applied, and if the creator diffusion indicator exceeds the saturation threshold, a saturation control operation is applied. In one embodiment, if the hook section start time is 25 seconds, the excess time relative to the breakout threshold is calculated as 10 seconds, and the damping operation may be applied.

[0072] Additionally, in this specification, the “saturation threshold” may be set based on the point where the slope of increase of the viral potential index relative to the increase of the creator diffusion index decreases below a reference value, and the “minimum slope” may be set as a system setting value or a statistical-based value as the reference value of the said slope of increase.

[0073] The above-described audio source-vocal compatibility index analysis module transmits a command to derive a compatibility index by quantifying physical frequency conflicts between vocal and accompaniment sources and degradation of clarity caused by excessive sound pressure. It automatically transmits a collection command to acquire short-time Fourier transform data at the audio buffer level and to measure the integrated loudness and dynamic range values ​​of the corresponding section through real-time metering. After calculating a base score by applying log correction to the identified masking overlap rate, if the ratio of loudness to dynamic range exceeds a preset overcompression threshold of 3.0, it transmits a control command to activate an exponential attenuation correction value to cause the score to drop sharply. If it is identified that the audio source is in an overcompressed state with a ratio of dynamic range to loudness of 4.1, it automatically transmits a calculation command to determine severe degradation of clarity by calculating a compatibility index of 32 points through exponential attenuation correction rather than simple subtraction when 8% frequency masking occurs. In parallel with this, the real-time parameter change tracking module continuously monitors the amount of change in the operator's arrangement and mixing operations and transmits a command to dynamically determine whether the screen refresh engine is operating. It automatically transmits a collection command to acquire the absolute value of the change in parameters, such as the user's equalizer or compressor, and to acquire the load indicator of the graphics rendering thread reported by the operating system. If the absolute value of the operation change falls below the minimum significant change threshold automatically set based on genre-specific statistical variance, it transmits a control command to activate closed-loop control that immediately blocks unnecessary screen refreshes and switches to an internal accumulation mode. In an electronic music project environment where the automatically calculated minimum significant change threshold is 1.50%, if the user performs fine operations at the 0.5% level, it automatically transmits a control command to block the screen refresh thread by determining that this is below the threshold, thereby preventing system utilization spikes and audio interruptions.

[0074] The above modification guide generation module transmits a command to suggest a practical timeline arrangement direction for improving musical quality based on the fluctuation results of the calculated indices. It automatically transmits a command to collect data on the expected fluctuation range of index scores and the probability of side effects occurring when manipulating specific parameters from the above success potential index calculation module and the audio-vocal compatibility index analysis module. It transmits a control command to sort the recommendation rankings for multiple guide candidate groups through an asymmetric compensation operation, which evaluates the expected profit (success index increase) gently by applying a logarithmic function, and evaluates the risk factor (intelligence degradation probability) harshly by applying an exponential function. It automatically transmits a calculation command to generate a top-priority proposal card by calculating the guide efficiency score of the 2.5 kHz band -1.5 dB cut operation, which was derived through virtual simulation as having an expected index increase of 3.2% and a side effect probability of 0.1%, as the first priority. Finally, the above information visualization output unit transmits a command to operate as a hardware display and graphic interface mediating unit that visually displays the system calculation results and the generated arrangement improvement guide to the user. It automatically transmits a communication command to receive text and score data from the above modification guide generation module and a screen rendering allow switch signal from the parameter change real-time tracking module. It updates and outputs the donut gauge and frequency heatmap only when the screen update allow signal is received, and transmits a control command to prevent visual interruption by performing forced synchronization once upon the expiration of the maximum delay timer when the cumulative calculation mode is prolonged. When the amount of change in the user's parameter manipulation exceeds the above threshold and an allow signal is received, it automatically transmits an output command to immediately render the guidance text generated as the first priority, stating "Previous week 10-second reduction needed," and the potential index graphic updated to 85 points in the top navigation area of ​​the screen.

[0075] This system evaluates the commercial value and acoustic completeness of a sound source in stages based on collected multimodal data, and the specific operation process is as follows.

[0076] First, the dynamic trend momentum calculation process applies a damping coefficient based on the elapsed collection period to behavioral indicators obtained from an external platform, and calculates dynamic trend momentum weights by applying a log correction operation when the ratio of the calculated behavioral indicators to market competition intensity exceeds an abuse detection threshold (10.0).

[0077] Next, the main source-based viral potential evaluation process calculates the start time of the hook segment through audio timeline analysis, applies a damping operation linked to the excess time if the start time exceeds the exit threshold (15 seconds), and calculates the viral potential index by applying a saturation control operation if the creator diffusion indicator exceeds the saturation threshold.

[0078] Next, the audio-vocal compatibility index determination process calculates the audio-vocal compatibility index using the masking overlap rate and clipping count between the vocal and accompaniment, and a combined indicator of the integrated loudness unit (LUFS) and dynamic range.

[0079] Additionally, in this specification, “Unified Loudness Unit (LUFS)” is a unit representing the perceptual sound pressure of a time interval, and “Decibel Full Scale (dBFS)” is a unit based on the maximum amplitude that a digital audio sample can express. The sound source-vocal compatibility index analysis module (150) can calculate the unified loudness and whether clipping occurs using the LUFS and dBFS standards.

[0080] Next, the system resource control process limits screen updates and switches to an internal accumulation method when the amount of change in user operation is less than the minimum significant change threshold (e.g., 1.50% for electronic music), and performs forced updates when the maximum delay timer expires.

[0081] The aforementioned trend weight learning module transmits a command to calculate dynamic trend momentum weights by quantifying the influence of collected external social trend keywords, short-form metadata, and listener behavior data on current market trends. The operation of this calculation logic automatically transmits control commands to derive reliable trend indicators by preemptively defending against information obsolescence caused by the short meme lifespan characteristic of short-form platforms and blocking data distortion caused by abuse by specific fandoms. The related computational processing procedure is structured based on two key axes—the temporal freshness of data and the intensity of market competition—and transmits evaluation commands to the internal engine. Specifically, based on the diffusion theory of social networks, which states that trend influence declines exponentially rather than linearly the further back the data was collected, it automatically transmits control commands to apply an exponential decay logic over time. Simultaneously, in addition to the economic principle that the scarcity of a trend decreases even if the behavior indicators of a specific track surge, if the indicators of other tracks released at the same time are also high, it transmits a command to apply a log-based penalty function to the ratio of behavior indicators to competition intensity.

[0082] The aforementioned trend weight learning module automatically transmits commands to periodically collect data through an external platform interface. For example, it transmits a collection command to acquire quantitative data in a manner where the daily usage of short-form challenge templates of a specific genre is acquired as 500,000, the collection time is 3 days prior to the present, and the average daily usage of the top 100 challenges during that period is collected as 800,000. The dynamic trend momentum weights calculated in this way are automatically transmitted to the success potential index calculation module. A computation instruction command is transmitted to utilize the transmitted weight values ​​as a principal multiplier applied to the cosine similarity between the feature vector of the input audio source and the trend reference vector of the market hit song cluster, thereby dynamically scaling the final viral potential index. As a specific processing procedure, if 5 days have passed since data collection, the system automatically transmits a command to exponentially attenuate the reflection ratio of the corresponding sample data to less than 50% according to a pre-set half-life rule. In addition, even if the social buzz volume of a specific song is 100,000, if the market average buzz volume calculated by the system is identified as 200,000, the relative dominance of the corresponding trend indicator is judged to be low, and a processing command is transmitted. If an excessive concentration situation is detected where a specific song records 1,000,000 while the market average buzz volume is 10,000, a command is automatically transmitted to filter distortion caused by artificial streaming by the fandom by applying a logarithmic function to the scale to mitigate it, rather than applying it directly proportionally to the behavioral indicator ratio. Finally, a command is transmitted to determine the final momentum weight by combining the aforementioned attenuated ratio and the corrected scale.

[0083] Additionally, in this specification, the “market hit song cluster” is a reference set for generating a trend reference vector, and in one embodiment, it may be defined as a set of sound sources that have existed within the top 100 of chart data for the last 12 months for 7 consecutive days or more, and the trend reference vector may be generated from the average or center value of the feature quantity calculated from the sound sources of the hit song cluster.

[0084] Regarding exception handling measures, if a situation is detected where it is impossible to obtain real-time competition intensity data due to communication failures of external platforms, the module automatically transmits a command to prevent the interruption of the calculation logic by inputting the moving average value of the last 30 days as an alternative competition intensity. In addition, to prevent the momentum weight from converging to mathematical zero due to exponential decay, it transmits a defensive command to preemptively prevent computational errors and zero splitting in the logic by setting a pre-configured lower threshold of 0.1. Unlike existing prior methods that fail to predict trends by applying simple weighted sums or linear proportions to specific spam keywords or trend data and judging 1,000,000 view data from one week after release as having the same value as 1,000,000 view data from yesterday, this system automatically transmits a thoroughly differentiated evaluation command. That is, according to the present invention, the value of 1,000,000 view data from one week ago is drastically reduced by exponential decay logic, and even if the data was generated yesterday, if the views of other sound sources at the same time are 2,000,000 views, a control command is transmitted so that the momentum is evaluated low by log correction. Through this, a command is automatically transmitted to extract only the actual trend influence into the database by simultaneously defending against information obsolescence and market overheating.

[0085] The results of multiple simulations conducted to verify the operation of this logic are also reflected within the system to transmit operation commands. First, if the elapsed time is 1 day, the behavioral indicator is 500,000 cases, and the competition intensity is 100,000 cases, the data ratio is calculated as 5.0; since this indicates high recency and low competition, a command is automatically transmitted to rapidly increase the momentum weight to 0.85. Conversely, if the elapsed time is 1 day and both the behavioral indicator and competition intensity are 500,000 cases, the data ratio becomes 1.0, so a command is transmitted to evaluate the momentum weight at an intermediate level of 0.50. In an extreme case where abuse is suspected, with the behavioral indicator at 5,000,000 cases and the competition intensity at 100,000 cases, a command is automatically transmitted to forcibly defend the momentum weight at a limit of 0.95 through log control, even though the data ratio explodes to 50.0. When 3 days have passed, a command is sent to reduce the weight to 0.65 to reflect the decline in freshness, even if the previous ratio was 5.0. When the elapsed time increases to 5 days, exponential decay begins in earnest, so a command is automatically sent to further lower the weight to 0.40. When 7 days have passed (one week), a command is sent to reduce the weight to 0.22, judging that the trend influence has been significantly lost. At this point, even if the absolute values ​​are high—such as 5,000,000 behavioral indicators and 500,000 competition intensity—if 7 days have passed, the weight is evaluated as low at 0.35, and a processing command is automatically sent. Furthermore, if 10 days have passed and the data ratio is 1.0, a command is sent to process it as having entered the extinction phase by determining the weight to be 0.15. When 14 days have passed, even if the ratio is 5.0, a command is automatically sent to forcibly fix the weight at a pre-set lower limit of 0.10 to prevent convergence to 0. Finally, even if the elapsed time is 1 day, if the behavior indicator is 10,000 cases and the competition intensity is 1,000,000 cases, and the competition is overwhelming, it is considered that the trend-following has failed, so a command is sent to process the weight as the lower limit value of 0.10.

[0086] Key judgment rules for system operation are also mapped to conditions and execution logic to automatically transmit configuration commands. Specifically, if the elapsed data collection period is 3 days or more, a command is transmitted to activate the daily exponential decay logic. If the ratio of behavioral indicators to competition intensity exceeds 10.0, a command is automatically transmitted to activate the log-based scale suppression logic for abuse filtering. If the reception of competition intensity data fails, a command is transmitted to activate the logic to process it by replacing it with the recent 30-day moving average data. If the momentum weight calculated as a result of the operation is less than 0.1, a command is automatically transmitted to execute the logic to forcibly fix the final momentum weight to the lower limit value of 0.1. Update commands are transmitted to allow the system to dynamically apply statistical inflection points derived from the pre-established short-form challenge lifecycle analysis model covering the past 5 years, such as the 3-day threshold for exponential decay activation and the ratio of 10.0 as the threshold for suppression logic. This automatically transmits commands that fully secure logical necessity and industrial objectivity by operating based on the point where the average diffusion cycle of a meme sharply declines and the threshold ratio that platform algorithms consider as abnormal traffic. Ultimately, the adoption of this logic for calculating dynamic trend momentum transmits commands that overcome the limitations of conventional technology, which simply accumulates past data and calculates an average. In other words, it automatically transmits driving commands that enable the system to immediately adapt to the rapidly changing trends of the short-form market and fundamentally prevent recommendation algorithms from being disrupted by streaming offensives from specific fandoms, thereby providing highly reliable success prediction indicators to managers and music producers.

[0087] The above trend weight learning module (120) may use the following data processing formula to calculate dynamic trend momentum weights.

[0088] [Operation Formula 1]

[0089] Dynamic Trend Momentum Weight = (Natural logarithm of the value obtained by dividing the behavioral indicator by competition intensity) × (Elapsed Days Damping Factor)

[0090] Here, the natural logarithm is a logarithm with base natural constant, and the elapsed day decay coefficient can be set to 1 when the collection elapsed day is 0 days, 0.5 when it is 5 days, and 0.25 when it is 10 days.

[0091] [Table 1-1] 10-Ratio Simulation Data for Setting Abuse Detection Threshold (Ratio 10.0)

[0092]

[0093] [Table 1-2] Simulation for Verification of Effect of Threshold (Ratio 10.0) Setting

[0094]

[0095] As shown in Table 1-1 above, when the output ratio reaches 10.0, a statistical inflection point is identified where the reliability of the actual market data drops to less than half, and an update command is transmitted to establish this as the judgment threshold. Accordingly, even if an abnormal situation is input where the ratio explodes to 20.0 as shown in Table 1-2, a control command is automatically transmitted to block indicator distortion by safely defending the momentum weight to a level of 2.99 through the log correction of [Equation 1] above.

[0096] Tables 1-1 and 1-2 above represent simulation examples for verifying whether a section in which the dynamic trend momentum weight increases excessively occurs as the output ratio increases. In one embodiment, the abuse detection threshold may be set based on a point where a section in which the range of weight fluctuation increases rapidly in proportion to the increase in the output ratio is observed, and may be configured so that a log correction operation is applied in a section exceeding the threshold to limit the range of weight fluctuation.

[0098] The aforementioned success potential index calculation module transmits a command to calculate a viral potential index that quantifies the probability of success in short-form and streaming environments by fusing the audio, lyrics, and structural characteristics of the input sound source. The purpose of the aforementioned judgment criteria is to automatically transmit control commands to provide a reference point for timeline arrangement optimization by excluding song selection methods that relied on the subjective listening of the person in charge, and by quantifying the initial churn rate and structural pleasure preferred by platform algorithms into objective indicators. The calculation procedure transmits computation commands composed of three axes: trend alignment, structural pleasure, and platform algorithm defense capability. Specifically, as a procedure, it automatically transmits a command to perform a first-order calculation of trend alignment through the multidimensional feature vector similarity between the input sound source and market hit songs. Subsequently, based on psychoacoustic grounds that attract the auditory attention of the public, it transmits a control command to adjust the index upward in proportion to the dynamic change amount and prosodic regularity at the beginning of the song. Furthermore, based on the principle of viewing retention rate, which is a core recommendation criterion of short-form algorithms, it automatically transmits a command to forcibly lower the index in an inverse relationship as the entry point into the hook section is delayed. In addition, when the creator participation diffusion indicator reaches an oversaturated state, an exponential saturation logic is applied to send a control command to suppress the score from increasing indefinitely.

[0099] The above module automatically transmits commands to acquire data by scanning the audio buffer of a digital audio workstation or analyzing input files. Specifically, it transmits a collection command to analyze the initial 60-second segment to extract the initial dynamic change from 0 seconds to the 10-second mark as a 4.5dB increase, and to scan the chorus entry point where audio energy surges to quantitatively acquire the hook start point in absolute time units such as 12 seconds. Additionally, it automatically transmits commands to collect creator diffusion indicators, such as the number of derivative user-generated content items for the song, by linking with a social platform application interface, such as 15,000 items. It transmits an output command to visualize the calculated viral potential index and the viral probability normalized to a range of 0 to 100% in real-time as a donut-shaped gauge on the global navigation header of a dashboard or plugin. Going beyond simply providing scores, it automatically transmits control commands to directly utilize these as judgment criteria for generating timeline arrangement optimization suggestion cards from an AI co-producer, such as increasing or decreasing specific frequency bands or shortening the intro. As part of the system's step-by-step processing procedure, a computation command is transmitted to calculate the similarity between the input audio feature vector and the trend reference vector as 0.75 and proportionally multiply it by the previously calculated dynamic trend momentum of 0.8 to secure a first base index of 0.60. If the analyzed initial dynamic change is high and the affricative impact of the lyrics is identified as strong, a control command is automatically transmitted to proportionally add a reward weight of 15% to the aforementioned first base index. If the hook start time is detected as 20 seconds, exceeding the pre-set short-form exit threshold of 15 seconds, a penalty computation command is transmitted to attenuate the index by 30% by applying an inverse formula where the denominator increases. If the creator diffusion indicator exceeds 50,000 instances, reflecting the phenomenon where algorithm exposure decreases due to accumulated public meme fatigue, the final viral potential index is set to 0.0 through an exponential saturation logic where the growth curve converges to 1.0.Automatically sends a command to confirm as 82.

[0100] In the event of an exceptional situation where a clear hook section is not detected because the input audio is of the ambient or classical genre, the system recognizes the spectral center variance flatness of the audio and sends a command to automatically switch the logic from short-form evaluation mode to pure streaming chill evaluation mode. Through this, it automatically sends a control command to block excessive inverse penalty calculations caused by the absence of a hook and to ensure computational stability by adding a constant to the denominator of the time variable to prevent zero splitting. To prevent the distortion of conventional technology, which uniformly assigns high success scores regardless of the song's structure simply because the tempo is fast and the energy feature is high, this system sends a command to perform audio timeline analysis. That is, regardless of how high the tempo and energy are, if it is identified that the hook start point is 40 seconds, the inverse penalty logic is immediately activated, and a strong suppression command is automatically sent to calculate a low viral potential. Based on the key judgment rules set within the system, if the hook start point is 15 seconds or less, a command is sent to exempt the inverse penalty and activate the initial dynamic compensation ratio. On the other hand, if the hook start time exceeds 15 seconds, a control command is automatically transmitted to apply a penalty for failure in the initial churn prevention rate by activating an inverse decay logic based on the excess amount per second. If the creator diffusion indicator exceeds the saturation threshold, an exponential saturation suppression logic is activated, and if an imbalance index between energy and sentiment is detected, a calculation command is transmitted to apply a negative correction weight based on the average churn rate by genre. The aforementioned 15-second hook start time threshold is not an arbitrary setting by the operator, but rather a statistical inflection point derived by analyzing average viewer churn rate log data from major short-form platforms is automatically updated by transmitting an update command.

[0101] Furthermore, the saturation control logic for the diffusion indicator transmits control commands to operate by reverse-engineering the actual platform's operational mechanism, where an algorithm forcibly applies diversity quotas to reduce exposure frequency when a specific audio source monopolizes the feed. Through the adoption of this process, the system automatically transmits output commands that go beyond simple evaluation to provide arrangement optimization navigation based on a clear causal relationship: shortening the intro by 5 seconds on the timeline brings the hook start point within 15 seconds, thereby releasing the inverse penalty and increasing viral potential by 12%. As a term and threshold standard for system operation, a command is transmitted to quantitatively calculate the initial dynamic variation amount in dB units, representing the maximum variation of the root mean square level between designated 10-second blocks within the initial 60 seconds of the audio timeline. For the hook start point, a control command is automatically transmitted to determine, in seconds, the absolute time at which spectral energy and prosodic density first form local maximums across the entire time axis of the song. Finally, the exponential saturation threshold is defined as a limit standard value based on a platform algorithm that controls the increase slope of the viral potential index calculation value to drop below a preset minimum slope when the creator diffusion indicator increases, and a command is sent to apply it to the entire system.

[0102] The above-mentioned box office potential index calculation module (140) may use the following data processing formula to calculate the viral potential index.

[0103] [Equation 2]

[0104] Viral Potential Index = (Feature Similarity × Dynamic Trend Momentum Weight) + Initial Dynamic Reward - (Overtime × Inverse Damping Factor)

[0105] Here, “excess time” is the number of excess seconds when the hook interval start time exceeds the deviation threshold (15 seconds), and the “inverse damping coefficient” can be set as the reciprocal of the value obtained by adding 1 to the excess time. For example, if the excess time is 0 seconds, the inverse damping coefficient is 1, and if the excess time is 10 seconds, the inverse damping coefficient can be set to 1 / 11.

[0106] [Table 2-1] 10 simulation data for setting the hook start threshold (15 seconds)

[0107]

[0108] [Table 2-2] Simulation for Verification of Effect of Threshold (15 seconds) Setting

[0109]

[0110] Based on actual data from Table 2-1 above, which shows that the platform's viewing retention rate drops to 50% or less when 15 seconds have passed since the hook started, a command to determine the threshold is automatically transmitted. Through this, when a song with a long intro set to 25 seconds is input as shown in Table 2-2, a damping constant proportional to the excess 10 seconds is applied through [Equation 2] above to lower the potential index to 45 points and a visual guidance command instructing to shorten the intro is transmitted.

[0111] Tables 2-1 and 2-2 above represent simulation examples to determine whether a decreasing trend in viewing duration indicators is observed as the hook section start time increases. In one embodiment, the break-out threshold (15 seconds) may be set based on the point where a section in which the decrease in viewing duration indicators widens is observed in the simulation.

[0113] The above-described audio-vocal compatibility index analysis module transmits a command to calculate the audio-vocal compatibility index by quantifying physical conflicts and clarity degradation across frequency bands between the vocal source and the accompaniment source. The purpose of the above judgment criteria is to automatically transmit control commands to exclude masking errors that occurred in conventional mixing environments due to reliance on the operator's subjective hearing, and to provide a standard for precise acoustic correction guidelines by calculating the degradation of audible clarity caused by over-compressed audio sources due to excessive compression and limiting into objective physical values. The calculation procedure consists of three stages of numerical computation based on psychoacoustic masking effects and audio signal processing theory, and transmits driving commands. Specifically, it automatically transmits a computation command to gradually reduce the basic compatibility index by applying a logarithmic function to the average masking overlap rate of the problem band identified through short-time Fourier transform-based frequency domain analysis. Subsequently, it transmits a control command to apply the ratio of digital clipping events occurring within the real-time audio buffer as a direct linear subtraction factor. Third, it calculates the ratio of loudness to the dynamic range of the input sound source and automatically transmits a command to rapidly attenuate the aforementioned compatibility index in the form of an exponential function as the ratio increases. This transmits a command to control the system by reflecting the acoustic fact that the lethality of frequency overlap on clarity in a state of insufficient headroom is non-linearly amplified.

[0114] The above module automatically transmits a command to collect audio sample data in real time from the master bus or individual track buffers of a digital audio workstation. It transmits a collection command to obtain a masking occurrence rate of 7.9% in the 237Hz band as a result of performing a Short-Time Fourier Transform on the vocal and accompaniment signals by dividing them into buffers of 1024 samples. Simultaneously, it automatically transmits a command to quantitatively obtain 15 clipping samples exceeding 0 dBFS within the buffer section, with the integrated loudness of the corresponding buffer section measured as -8 LUFS and the dynamic range as 3dB via a real-time audio meter. It transmits an output command to display the calculated compatibility index score of 88.2 points and the 237Hz frequency masking area on the user interface as an explainable AI-based visualization heatmap. This result is transmitted to a modification guide generation module, and a control command is automatically transmitted to utilize it as a key judgment criterion, providing the creator with a timeline arrangement optimization suggestion card in a navigation format for clarity improvement, such as the need for -3dB attenuation in the 237Hz band. As a specific scenario of the processing procedure, if 7.9% overlap between the vocal and accompaniment signals is identified in the 237Hz band, the system transmits a command to perform a preliminary calculation of a reference compatibility index of 88.2 points by applying a preset log attenuation formula. Subsequently, if 15 instances of clipping are detected within a 1024-sample audio buffer, a calculation command is automatically transmitted to subtract -1.5 points—a penalty constant proportional to the clipping ratio of 15 / 1024—to arrive at a score of 86.7 points. If the sound pressure in the corresponding section is identified as being in an excessively compressed state with -8 LUFS and a dynamic range of 3dB, it is determined that the masking impact is more severe than in the case of simple overlap; therefore, a control command is transmitted to activate exponential attenuation logic to drastically lower the score from 86.7 points to a final score of 71.5 points.

[0115] An exception handling command is automatically transmitted to prevent calculation errors that occur when the vocal energy, which serves as the denominator, becomes zero when the user mutes the vocal track in the audio editing program or plays an instrumental section where vocals have not yet been recorded. When the root mean square level of the vocal track in the audio buffer is detected to be below -80dB, which is a preset noise floor, the module immediately bypasses the compatibility index calculation logic and switches to an instrumental-only mode to prevent system crashes and transmits a control command to block the application of unnecessary penalties. To overcome the limitations of prior art, where a general spectrum analyzer provides only visual information regarding overlapping frequency bands and cannot quantify clarity degradation, the present invention automatically transmits a differentiated evaluation command. That is, even if the same 7.9% masking occurs in the 237Hz band, the system independently determines whether the section is an acoustic section with a wide dynamic range or an over-compressed section; if it is the latter, it applies an exponential penalty and transmits a calculation command that converts the muffled sound actually perceived by the ear into an accurate physical score. As a primary judgment rule, if the masking overlap rate of a specific band exceeds 5%, a command is automatically sent to activate the corresponding frequency band visualization heatmap and apply log-based basic attenuation. If clipping exceeding 0 dBFS occurs within the buffer, a command is sent to activate direct subtraction logic proportional to the ratio of clipping samples. If the dynamic range to loudness ratio exceeds a threshold, an exponential attenuation penalty is activated due to the risk of psychoacoustic clarity degradation, and a control command is automatically sent to forcibly stop the operation if the root mean square level of the vocal track is less than -80dB.

[0116] The threshold for the dynamic range-to-loudness ratio that triggers the aforementioned exponential attenuation is not a value arbitrarily designated by the operator; instead, it transmits a command to update the reference value dynamically applied by the system by database-ing the -14 LUFS target—the official loudness specification required by global streaming platforms—and audio engineering standard mastering guidelines. Therefore, control commands are automatically transmitted to ensure that the application of frequency masking penalties in over-compressed states possesses a technical necessity for complying with global music distribution standards. Through the adoption of this calculation logic, when a creator excessively raises the sound pressure of a song, the system immediately detects the dynamics ratio and vertically drops the vocal compatibility score, thereby transmitting an output command that visualizes the trade-off between securing sound pressure and vocal clarity as a real-time numerical value. Consequently, it automatically transmits driving commands that guide even non-mixing experts to achieve commercial sound quality compliant with platform standards. As a criterion for system control, the masking overlap rate is calculated by transmitting a command based on the ratio of the area where the energies of the vocal and accompaniment signals coexist at the frequency level—derived through short-time Fourier transform analysis—and intersect beyond a pre-set masking threshold curve. A digital clipping event automatically transmits a control command to determine that the waveform's peaks are clipped when the quantized digital audio sample value reaches or exceeds 0 dBFS, which is the maximum amplitude the system can represent. Finally, a command is transmitted to apply a dynamic reference value to the system logic, defined as the threshold for the ratio of loudness to dynamic range, which is the absolute value of the input sound pressure of the input source divided by the dynamic range, which is the difference between the maximum peak and the average level of the corresponding section; the larger this value, the more excessively the audio signal is compressed.

[0117] The above-mentioned sound source-vocal compatibility index analysis module (150) may use the following data processing formula to calculate the sound source-vocal compatibility index.

[0118] [Equation 3]

[0119] Audio-Vocal Compatibility Index = Base Score - (Log-based Masking Overlap Correction) - (Clipping Count × Linear Subtraction Constant) - (Exponential Attenuation Correction)

[0120] Here, the exponential damping correction value can be determined by a combined indicator of the “absolute value of integrated loudness” and the “dynamic range,” and the exponential damping correction value can be configured to be applied when the combined indicator exceeds an overcompression threshold (e.g., 3.0).

[0121] [Table 3-1] 10-Round Simulation Data for Setting Overcompression Detection Threshold (Ratio 3.0)

[0122]

[0123] [Table 3-2] Simulation for Verification of Effect of Threshold (Ratio 3.0) Setting

[0124]

[0125] As described in Table 3-1 above, when the calculation ratio of loudness relative to dynamic range exceeds 3.0, a command is transmitted to set the value as an over-compression detection threshold, reflecting the phenomenon where the clarity evaluation score perceived by the human ear collapses to 40 points or less. Based on this, if an over-compressed sound source with a ratio of 4.1 is identified as shown in Table 3-2, the exponential attenuation correction value of [Equation 3] above is strongly activated to vertically drop the final index to 32 points, and a warning output command requesting limiter relaxation is automatically transmitted.

[0126] Tables 3-1 and 3-2 above represent simulation examples to determine whether a tendency for the decrease in the compatibility index to expand is observed in the interval where the coupling index exceeds a threshold. In one embodiment, the over-compression threshold may be set based on the point where the interval in which the decrease in the compatibility index expands with increasing coupling index is observed in the simulation.

[0128] The aforementioned parameter change real-time tracking module and modification guide generation module calculate the efficiency of the system's guide provision based on the creator's editing operations and transmit commands to dynamically control whether the user interface is updated. The purpose of the aforementioned judgment criteria is to automatically transmit control commands to ensure an uninterrupted audio monitoring environment by preventing audio dropouts caused by resource contention between the audio signal processing thread and the screen rendering thread within the digital audio workstation environment, and by blocking unnecessary screen flickering. The aforementioned control procedure transmits driving commands by combining an asymmetric compensation mechanism and a closed-loop control method. Specifically, it automatically transmits a computation command to calculate the recommendation priority of the guide through asymmetric operations, which evaluate the expected increase in commercial success potential upon guide application by taking a logarithm to evaluate it gently, and evaluate the deterioration of the audio source-vocal compatibility index by applying an exponential function to evaluate it harshly. Subsequently, using the parameter change signal as a trigger, it transmits a control command to determine whether the absolute value of the change caused by user operation exceeds the minimum significant change amount defined by the algorithm's resonance threshold. If it falls below the threshold, it automatically transmits a command to immediately stop screen updates and switch to an internal accumulation mode.

[0129] The above module transmits commands to collect real-time data through a plugin host application interface. For example, it automatically transmits a collection command to acquire the amount of parameter change in real-time in 0.1dB increments when a user finely manipulates an equalizer in the 2.5kHz frequency band with a mouse. Simultaneously, it transmits a calculation command to measure the current audio engine's system allocation rate as a screen load indicator of 85% and quantitatively derive the absolute value of the expected viral potential index change based on the changed parameters as 0.8%. It automatically transmits a control command to utilize the calculated guide efficiency indicator as a reference value for a ranking algorithm that displays the safest and most effective card among dozens of arrangement improvement suggestion cards at the top of the navigation. Additionally, it transmits a command to utilize the result of determining whether the threshold is passed as a dynamic switch to decide whether to allow or delay the operation of the screen rendering engine. As a step-by-step processing procedure of the system, when the system simulates a 2.5kHz band -1.5dB cut guide and the expected viral potential index increase is 3.2% and the probability of deterioration of the source-vocal compatibility index is 0.1%, it automatically transmits a command to determine the efficiency indicator of the guide as the first priority through asymmetric compensation calculation.

[0130] When a user changes parameters by turning a knob with a mouse, a scanning command is transmitted when the absolute value of the expected change in the viral potential index is detected to be 0.8%. In a situation where the previously calculated minimum significant change is 1.5%, if the absolute value of the change occurring is determined to be below the reference threshold of 0.8%, the system automatically transmits a control command to prevent unnecessary flickering by adding the corresponding change to the internal accumulated value without refreshing the screen. Subsequently, the moment additional operations accumulate and the absolute value of the change exceeds the reference threshold of 1.5%, a command is transmitted to dynamically update the navigation path by recalculating the distance from the entire trend vector. Since a synchronization mismatch state may occur where the internal data is changed but the screen remains frozen if the user sustains fine operations below the threshold for a very long time, an exception handling command is automatically transmitted to prevent this. The module activates a pre-set maximum delay timer of 1.5 seconds; even if the change is below the minimum significant change, a control command is transmitted to prevent the interruption of visual feedback of the operation by forcibly refreshing the user interface once upon the expiration of the timer. To overcome the limitations of prior art, which attempts to immediately update the screen and score whenever the mouse moves, thereby causing system utilization spikes and audio interruption noise, this system automatically transmits commands that perform closed-loop control based on algorithmic resonance thresholds. That is, in fine manipulation sections that do not cause significant changes to the audio's potential for success, it blocks screen rendering resources and yields them to the audio processing thread, thereby dramatically reducing the load while transmitting control commands that provide key arrangement guides in a timely manner.

[0131] As a key judgment rule for controlling system resources, if the absolute value of the change in the viral potential index exceeds the minimum significant change, a command is automatically transmitted to dynamically update the navigation path and allow screen output. Conversely, if the absolute value is less than the minimum significant change, a command is transmitted to block screen updates and switch to internal accumulation mode. If the screen load indicator exceeds the system limit, a control command is automatically transmitted to delay guide screen output and extend the data update cycle to the maximum; if the accumulation operation mode time exceeds the maximum delay timer, a command is automatically transmitted to perform forced screen synchronization once. The minimum significant change threshold determining whether to allow screen updates is not a fixed constant arbitrarily designated by the operator, but rather an update command is transmitted to apply it to the system by calculating it from the average index variance within the genre-specific hit song clusters of the pre-trained trend weight learning module. Specifically, the threshold is automatically calculated as a statistical value corresponding to the square root of one-tenth of the variance value, and a control command is automatically transmitted to dynamically vary it based on the statistical significance of the tone changes required for each genre. By adopting this dynamic load control process, it maintains application stability while performing advanced 4D data and audio operations in real time, and transmits output commands that induce the navigation screen to react dynamically only when the audio arrangement work traces a trajectory that has a significant impact on viral potential. As a term standard for system control, it automatically transmits a command to apply the algorithm resonance threshold as a criterion value to determine whether a change in audio parameters caused by a single user edit operation results in a significant change in the success index. It transmits a calculation command by defining the minimum significant change amount as the result of a threshold formula derived based on the variance statistics of a pre-learned hit song cluster group.Screen load indicators are determined by the drop in frame rate per second and system utilization of the graphics rendering threads reported by the operating system, and control commands are automatically transmitted to be utilized for dynamic control.

[0132] The above parameter variation real-time tracking module (160) may use the following data processing formula to calculate the minimum significant change amount threshold.

[0133] [Equation 4]

[0134] Minimum significant change threshold = √(Variance of hit song group potential index / 10)

[0135] Here, “variance of the hit song group potential index” can be defined as the variance of the viral potential index of a hit song cluster by genre (e.g., last 12 months, top 100, 7 consecutive days or more), and “ ” can be set as a normalization factor to conservatively reduce the variance value.

[0136] [Table 4-1] 10 simulation data for automatic setting of minimum significant change threshold

[0137]

[0138] [Table 4-2] Simulation for Verification of Screen Load Control Effects on Dynamic Threshold Settings

[0139]

[0140] As shown in Table 4-1 above, the system transmits a calculation command to independently and automatically derive a minimum significant change threshold from genre-specific statistical data through [Equation 4] above without manual input by the operator. Based on this, when a user performs a fine operation of 0.5% in an electronic music project environment, as shown in Table 4-2, the system immediately determines that the change amount is less than the calculated threshold of 1.50% and automatically transmits a dynamic control command to completely prevent audio interruption by blocking the update of the screen rendering thread.

[0141] Tables 4-1 and 4-2 above illustrate examples of operations in which a minimum significant change threshold calculated from genre-specific statistical data is applied to screen refresh control. In one embodiment, when the amount of change in user operation in an electronic music project environment is 0.5% and the calculated minimum significant change threshold is 1.50%, the amount of change in operation is determined to be less than the threshold, so screen refresh may be restricted, and accordingly, the frequency of audio interruption may be reduced.

[0143] In one embodiment, an embodiment in which the system components and processes of the present invention are operated integrally is as follows.

[0144] The above audio source input unit automatically transmits a command to collect audio sample buffer data and timeline playback position information in real time from an electronic music genre project being worked on by a creator on a digital audio workstation. In parallel, the above trend weight learning module acquires concurrent challenge template usage data from an external social media platform, identifies situations where the ratio of behavioral indicators to competition intensity exceeds a threshold of 10.0, and transmits a computation command to defend and calculate a trend momentum weight of 2.99 through log correction. The above key source feature extraction module automatically transmits a collection command to spectrally analyze the collected audio waveform to derive an initial dynamic change amount with an increase of 4.5 dB, identifies that the start point of the main hook section is at the 25-second position, and stores it as a feature vector. Subsequently, the above popularity potential index calculation module transmits a driving command to calculate a viral potential index by combining the calculated trend momentum weight and the audio feature vector. At this time, it is determined that the 25-second hook start time analyzed above has exceeded the 15-second dropout threshold based on the pre-set platform viewing retention rate by 10 seconds, and a damping constant proportional to the excess is activated to automatically transmit a control command to recalculate the viral potential index, which was initially calculated as 90 points, downgraded to 45 points.

[0145] Simultaneously, the audio-vocal compatibility index analysis module transmits a computational command to identify an 8% frequency masking overlap between the vocal and accompaniment sources in the 237 Hz band and determine that the dynamic range-to-loudness ratio in that section is 4.1, exceeding the overcompression threshold of 3.0. Based on the overcompression determination result, it automatically transmits a control command to calculate the audio-vocal compatibility index as 32 points, indicating a severely degraded clarity state, by strongly applying an exponential attenuation correction value rather than a simple linear subtraction. The correction guide generation module transmits a computational command to generate an arrangement improvement proposal card instructing a 10-second reduction in the intro and a relaxation of the mastering limiter to recover the downgraded viral potential index and clarity degradation index. The module automatically transmits a command to determine the guide efficiency score of the proposal card as the top priority through an asymmetric compensation operation that applies a logarithmic function to the expected index increase and an exponential function to the side effect probability.

[0146] In accordance with these instructions, when the user fine-tunes the equalizer parameter in the 2.5 kHz band to a level of 0.5%, the parameter change real-time tracking module transmits a collection command to immediately scan the amount of change in operation. The tracking module automatically transmits a control command to determine that the amount of change in operation is less than the threshold by comparing the amount of change in operation of 0.5% with the threshold of 1.50%, which is the minimum significant change threshold automatically derived from the variance value of the hit song group. Based on the determination result, the tracking module transmits a closed-loop control command to immediately block screen updates and switch the parameter change amount to an internal accumulation mode to prevent audio interruption. Subsequently, if additional operations accumulate and the absolute value of the amount of change in operation exceeds the threshold of 1.50% or if a pre-set maximum delay timer of 1.5 seconds expires, a control command is automatically transmitted to the information visualization output unit to transmit a screen update allow switch signal. The information visualization output unit that receives the above permission signal automatically transmits an output command that provides the creator with an uninterrupted audio environment and timely visual feedback while minimizing system load by rendering the arrangement improvement guide text generated as a first priority and the dynamically updated potential index graphic in the top area of ​​the display.

[0147] Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. Various modifications are possible by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention. Explanation of the symbols

[0149] Audio source input unit (110) Trend weight learning module (120) Main source feature extraction module (130) Box office potential index calculation module (140) Sound source-vocal compatibility index analysis module (150) Parameter change real-time tracking module (160) Modification guide generation module (170) Information visualization output unit (180)

Claims

Claim 1 An audio source input unit that receives an audio signal subject to multimodal analysis from a digital audio workstation or external storage medium, and receives audio sample buffer and timeline information of a master bus or submix bus within a project via a plugin extension protocol of the digital audio workstation; a trend weight learning module that calculates dynamic trend momentum weights of the market by combining reaction data from social media and short-form platforms; a key source feature extraction module that performs frequency analysis on the audio waveform received from the audio source input unit and calculates the initial dynamic change amount and the hook section start time; a lyric feature extraction module that derives lyric features including the number of syllables, the ratio of repeated phrases, and the ratio of plosive-family consonants and vowels from the lyric text corresponding to the audio signal; a feature fusion module that generates fused features by normalizing and combining the key source features and the lyric features; a success potential index calculation module that calculates the viral potential index of a song (a form of success potential index) using the dynamic trend momentum weights and the fused features; and the degradation of clarity caused by physical frequency collision between vocal and accompaniment sources and excessive sound pressure An audio-vocal compatibility index analysis module that quantifies and derives an audio-vocal compatibility index; a parameter variation real-time tracking module that continuously monitors changes in the operator's arrangement and mixing operations to dynamically determine whether the screen refresh engine is operating, and obtains changes in equalizer or compressor parameter operations as absolute values;A system for calculating music source success potential and providing real-time arrangement optimization guides through AI-based multimodal data fusion, comprising: a modification guide generation module that collects data on the expected index score fluctuation range and the probability of side effects occurring when specific parameters are manipulated from the above-mentioned success potential index calculation module and the above-mentioned audio-vocal compatibility index analysis module, sorts recommendation rankings for multiple guide candidate groups through asymmetric compensation operations in which the increase in the success index (expected profit) is evaluated gently by applying a logarithmic function and the probability of clarity impairment (risk factor) is evaluated harshly by applying an exponential function, and generates guide cards including timeline adjustment and sound adjustment guides; and an information visualization output unit that visually outputs the guide cards and index display information according to a screen refresh allow signal from the above-mentioned parameter fluctuation real-time tracking module. Claim 2 A system for calculating music success potential and providing real-time arrangement optimization guides through AI-based multimodal data fusion, characterized in that, in claim 1, the trend weight learning module collects behavioral indicators including the usage of challenge templates for a specific song from an external platform, data collection elapsed days, and market competition intensity data for the same period, applies an exponential decay logic over time based on the collection elapsed days, determines whether the calculation ratio of the behavioral indicator relative to the collected market competition intensity exceeds a preset abuse detection threshold, and if the calculation ratio is determined to exceed the abuse detection threshold, applies a correction operation based on a logarithmic function to the calculation ratio to limit the fluctuation range of the dynamic trend momentum weight, thereby mitigating indicator deviations caused by short-term traffic concentration of a specific group. Claim 3 In paragraph 2, the abuse detection threshold is set to a constant value of 10.0, and the trend weight learning module drives the calculation logic by inputting the moving average value of the last 30 days as a substitute competition strength to prevent system errors when an exceptional situation is detected in which the reception of the market competition strength data fails, and if the dynamic trend momentum weight calculated as a result of applying the exponential decay logic is derived to be less than 0.1, it is limited to a lower limit value of 0.1 to prevent zero division operation errors, characterized by an AI-based multimodal data fusion system for calculating music success potential and providing real-time arrangement optimization guides. Claim 4 In claim 1, the main source feature extraction module calculates the square root mean square level change range at the initial 10-second point relative to the start point of the audio signal waveform in decibels to derive the initial dynamic change amount, and determines the start point of the hook section where the spectral energy forms a local maximum value in absolute time units; the success potential index calculation module adds a compensation value proportional to the initial dynamic change amount to the value obtained by multiplying the similarity between the feature vector of the input sound source and the trend reference vector by the dynamic trend momentum weighting, determines whether the determined start point of the hook section exceeds 15 seconds, which is a dropout threshold set based on average viewer dropout rate data of a short-form platform, and if the determination result exceeds 15 seconds, applies a decay operation linked to the time exceeding 15 seconds to adjust the viral potential index, thereby generating a timeline adjustment guide including intro shortening, etc., thereby providing a guide for calculating sound source success potential and real-time arrangement optimization through AI-based multimodal data fusion. System. Claim 5 In claim 4, the above-mentioned success potential index calculation module further collects creator diffusion indicators including the number of user-generated content generated from the audio signal from an external social platform, continuously monitors whether the collected creator diffusion indicators exceed a preset saturation threshold, and if the monitoring result determines that the saturation threshold is exceeded, applies a saturation control operation to limit the increase of the viral potential index within a certain range, characterized by an AI-based multimodal data fusion system for calculating music success potential and providing a real-time arrangement optimization guide. Claim 6 In claim 4, the above-mentioned success potential index calculation module calculates the spectral center variance flatness of the audio signal to internally identify an exceptional situation in which the hook section, which is a distinct energy surge point, is not detected because the audio signal corresponds to an ambient genre or a classical genre, and if the calculated flatness satisfies a preset short-form evaluation exclusion criterion, switches to a streaming evaluation mode without applying attenuation calculations due to exceeding 15 seconds from the start of the hook section, thereby mitigating the application of excessive attenuation due to the absence of the hook section, characterized by an AI-based multimodal data fusion system for calculating music success potential and providing a real-time arrangement optimization guide.

Images