Music generation method, apparatus and device, and readable storage medium
Through the depth model REMI generation and analysis of texture data, combined with the reference of real music, the texture structure is automatically updated and combined, which solves the problem of relying on experts and high costs in traditional music creation, and achieves efficient, innovative and harmonious music generation.
Patent Information
- Application Number
- CN202411269883.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-11
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-09-11
AI Technical Summary
Relying on music arrangement experts in the creation of traditional music results in high labor costs and high creative thresholds, and the generation of texture templates takes a long time and is highly dependent.
Depth model REMI is used to generate multiple texture data. By obtaining the reference texture of chord sequences and real music, analyzing and updating the texture structure, generating the target texture, and combining them to generate new music, reducing manual intervention.
It reduces the labor cost of music creation, improves the innovation and harmony of the music, reduces homogeneity and mechanical listening, and enhances the audience's sense of identity and resonance.
Smart Images

Figure CN119028301A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and more specifically, to a music generation method, device, equipment and readable storage medium. Background Art
[0002] Listening to music is a common form of leisure and entertainment, while traditional music creation mainly relies on music producers.
[0003] In the process of creating music, music producers need to have rich music theory knowledge, which leads to a high threshold for music creation. In order to solve the above problem, the prior art provides a method for generating new textures based on texture templates produced by music arranger experts, and generating new music according to the new textures. However, the generation of texture templates requires music arranger experts to understand the textures of different types of music and the key points of different types of music textures, which is time-consuming, has high labor costs, and is highly dependent on music arranger experts. Summary of the invention
[0004] In view of this, the present application provides a music generation method, device, equipment and readable storage medium, which are used to solve the disadvantage of high labor cost of generating new textures in the prior art.
[0005] In order to achieve the above objectives, the proposed solution is as follows:
[0006] A music generation method, comprising:
[0007] Acquire a chord sequence, a reference texture of a target real music piece, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music piece, and the reference texture is composed of a plurality of texture structures;
[0008] For each texture structure of the reference texture, the texture structure is analyzed, and the texture data with the highest matching with the texture structure among the texture data is updated to generate a target texture corresponding to the texture structure;
[0009] Combining the target textures to generate a music texture for performing the chord sequence;
[0010] A music piece is generated based on the music piece texture and the chord sequence.
[0011] Optionally, multiple texture data generated based on the deep model REMI, including:
[0012] Inputting text prompt parameters, midi data marked with text prompt parameters or random seeds into the REMI multiple times to obtain multiple output data;
[0013] Each output data is cut based on a preset time value threshold set to obtain multiple texture data.
[0014] Optionally, obtain a reference texture of the target real music, including:
[0015] Determine the target music type to be generated;
[0016] Selecting a real music piece that matches the target music piece type from a plurality of real music pieces as a target real music piece;
[0017] The texture of the target real music piece is used as a reference texture.
[0018] Optionally, the target music type is composed of any one or more of a target music style, a target music mood, a target music segment type and a target bpm.
[0019] Optionally, analyzing the texture structure, updating the texture data with the highest matching with the texture structure among various texture data, and generating a target texture corresponding to the texture structure, comprises:
[0020] Determining a first music attribute of the texture structure and a second music attribute of each texture data;
[0021] Selecting a target music attribute having the highest similarity to the first music attribute from each second music attribute, wherein the texture data corresponding to the target music attribute is the texture data having the highest matching degree with the texture structure;
[0022] Based on the deductive method of the texture structure, the texture data with the highest matching to the texture structure is updated to form a target texture corresponding to the texture structure.
[0023] Optionally, determining the first music attribute of the texture structure includes:
[0024] Determining a first duration, a first density, a first stability, a first rhythm, a first paragraph trend, a first pitch width and a first range of the texture structure;
[0025] Among them, the first duration, the first density, the first stability, the first rhythm, the first paragraph trend, the first pitch width and the first range of sound constitute the first music attribute.
[0026] Optionally, the deductive method based on the texture structure updates the texture data with the highest matching with the texture structure to form a target texture corresponding to the texture structure, including:
[0027] Based on the musical note interpretation mode of the texture structure, the musical note interpretation mode of the texture data that has the highest matching with the texture structure is adjusted, and after the adjustment, a target texture corresponding to the texture structure is obtained.
[0028] A music texture generating device, comprising:
[0029] An acquisition module, used to acquire a chord sequence, a reference texture of a target real music, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music, and the reference texture is composed of a plurality of texture structures;
[0030] An analysis module, for analyzing each texture structure of the reference texture, updating the texture data with the highest matching to the texture structure among the texture data, and generating a target texture corresponding to the texture structure;
[0031] A combination module, used for combining various target textures to generate a music texture for performing the chord sequence;
[0032] A generation module is used to generate music based on the music texture and the chord sequence.
[0033] A music texture generating device comprises a memory and a processor;
[0034] The memory is used to store programs;
[0035] The processor is used to execute the program to implement the various steps of the above-mentioned music generation method.
[0036] A readable storage medium stores a computer program thereon, wherein when the computer program is executed by a processor, each step of the above-mentioned music generation method is implemented.
[0037] It can be seen from the above technical solutions that the music generation method provided by the present application can obtain a chord sequence, a reference texture of a target real music, and a plurality of texture data generated based on the deep model REMI, wherein the reference texture is used to characterize the chord interpretation method of the target real music, and the reference texture is composed of a plurality of texture structures; for each texture structure of the reference texture, the texture structure is analyzed, and the texture data with the highest matching with the texture structure in each texture data is updated to generate a target texture corresponding to the texture structure; based on this, the target texture of the present application is the data obtained after the texture data generated by REMI is updated, and the texture data generated by REMI is highly random, thereby improving the innovation of the target texture and avoiding the homogeneity and mechanical listening of the music texture; at the same time, when generating the target texture, the texture of the real song is used as a reference, so as to learn the note arrangement method of the real texture, and apply the note arrangement method of the real texture to the texture data generated by REMI, thereby ensuring the harmony, hierarchy and audience recognition of the target texture. At the same time, the target real song is easy to obtain, and does not need to be provided by the arrangement expert, and the texture data generated by REMI does not need to be provided by the arrangement expert, which reduces the labor cost; then, each target texture can be combined to generate a music texture for interpreting the chord sequence; based on the music texture and the chord sequence, the music is generated. On this basis, the present application can combine the music texture and the chord sequence to form a new music, and complete the construction of the new music. At the same time, since the textures referenced by each target texture combined by the present application belong to the same reference texture, the adaptability of each target texture is strong, thereby improving the progressiveness, coordination and audience resonance of the music texture obtained by combining each target texture, and further improving the hierarchy and harmony of the chord sequence interpretation. It can be seen that the present application can adjust the texture data generated by REMI with reference to the real texture, generate the music texture, and combine the music texture with the chord sequence to form a new music. Thus, in the process of constructing a new music, the labor cost is reduced and the expert dependence in the music creation process is reduced. At the same time, this application can also improve the innovation of the newly generated music while ensuring the roundness, layering and audience recognition of the newly generated music. BRIEF DESCRIPTION OF THE DRAWINGS
[0038] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0039] Figure 1 A flow chart of a music generation method disclosed in an embodiment of the present application;
[0040] Figure 2 A structural block diagram of a music generating device disclosed in an embodiment of the present application;
[0041] Figure 3 This is a hardware structure block diagram of a music generating device disclosed in an embodiment of the present application. DETAILED DESCRIPTION
[0042] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0043] The present application embodiment provides a music generation method, which can be applied to various music creation systems or music communication systems, and can also be applied to various computer terminals or smart terminals. The execution subject can be a processor or server of the computer terminal or smart terminal. The method flow chart of the method is as follows: Figure 1 As shown, specifically including:
[0044] Step S1, obtaining a chord sequence, a reference texture of a target real music piece, and a plurality of texture data generated based on a deep model REMI.
[0045] Specifically, the chord sequence may include a combination of multiple target chords, and each target chord may be obtained in a variety of ways. For example, in response to a user's chord specifying operation, a user-specified chord may be used as a target chord; or a chord recommended by a chord generation system may be used as a target chord.
[0046] A target real music piece may be selected from various authorized songs, and the texture of the target real music piece may be obtained as a reference texture.
[0047] The target real music may be selected in a variety of ways. For example, in response to a user's designation, a designated song may be selected from various authorized songs as the target real music. Alternatively, the target real music may be selected from various authorized songs based on actual target music type requirements.
[0048] The reference texture is used to represent the instrument playing method and chord interpretation method corresponding to the target real music.
[0049] The reference texture is composed of multiple texture structures.
[0050] Different ways of playing musical instruments have different effects on the audience's emotions.
[0051] Each texture data may be composed of a plurality of single tone data for constructing chords.
[0052] Each single tone data may include pitch, loudness, note start time and note duration.
[0053] Each texture data may contain necessary structures for forming a chord, such as a root note, a third note, a fifth note, and the like.
[0054] REMI is an open source model, and the texture distribution of its output data lacks coordination. Therefore, the output data needs to be processed to obtain the texture data.
[0055] Step S2: for each texture structure of the reference texture, analyze the texture structure, update the texture data with the highest matching with the texture structure among the texture data, and generate a target texture corresponding to the texture structure.
[0056] Specifically, a machine algorithm can be used to learn each texture structure and update the matching texture data to obtain the target texture after the update.
[0057] Step S3: combine the target textures to generate a music texture for performing the chord sequence.
[0058] Specifically, each target texture may be combined according to the combination mode of each texture structure in the reference texture to generate a music texture.
[0059] The music texture can be a midi file, a WAV file or an MP3 file.
[0060] Step S4: Generate music based on the music texture and the chord sequence.
[0061] Specifically, the chord sequence can be played according to the music texture to form a music.
[0062] It can be seen from the above technical solutions that the music generation method provided by the embodiment of the present application can obtain a chord sequence, a reference texture of a target real music, and a plurality of texture data generated based on the deep model REMI, wherein the reference texture is used to characterize the chord interpretation method of the target real music, and the reference texture is composed of a plurality of texture structures; for each texture structure of the reference texture, the texture structure is analyzed, and the texture data with the highest matching with the texture structure in each texture data is updated to generate a target texture corresponding to the texture structure; based on this, the target texture of the present application is the data obtained after the texture data generated by REMI is updated, and the texture data generated by REMI is highly random, thereby improving the innovation of the target texture and avoiding the homogeneity and mechanical listening of the music texture; at the same time, when generating the target texture, the texture of the real song is used as a reference, so as to learn the note arrangement method of the real texture, and apply the note arrangement method of the real texture to the texture data generated by REMI, thereby ensuring the harmony, hierarchy and audience recognition of the target texture. At the same time, the target real song is easy to obtain, and does not need to be provided by the arrangement expert, and the texture data generated by REMI does not need to be provided by the arrangement expert, which reduces the labor cost; then, each target texture can be combined to generate a music texture for interpreting the chord sequence; based on the music texture and the chord sequence, the music is generated. On this basis, the present application can combine the music texture and the chord sequence to form a new music, and complete the construction of the new music. At the same time, since the textures referenced by each target texture combined by the present application belong to the same reference texture, the adaptability of each target texture is strong, thereby improving the progressiveness, coordination and audience resonance of the music texture obtained by combining each target texture, and further improving the hierarchy and harmony of the chord sequence interpretation. It can be seen that the present application can adjust the texture data generated by REMI with reference to the real texture, generate the music texture, and combine the music texture with the chord sequence to form a new music. Thus, in the process of constructing a new music, the labor cost is reduced and the expert dependence in the music creation process is reduced. At the same time, this application can also improve the innovation of the newly generated music while ensuring the roundness, layering and audience recognition of the newly generated music.
[0063] In some embodiments of the present application, the process of obtaining multiple texture data generated based on the deep model REMI in step S1 is described in detail, and the steps are as follows:
[0064] S10, inputting text prompt parameters, midi data marked with text prompt parameters or random seeds into the REMI multiple times to obtain multiple output data.
[0065] Specifically, data can be input into REMI repeatedly to obtain output data;
[0066] The data input each time may be a text prompt parameter, midi data marked with a text prompt parameter, or a random seed.
[0067] Text hint parameters can be used to characterize texture type requirements.
[0068] The duration of the midi data can be set according to actual needs, for example, it can be set to 3 minutes.
[0069] Random seeds can be randomly generated parameters.
[0070] S11. Cut each output data based on a preset time value threshold set to obtain a plurality of texture data.
[0071] Specifically, the time value threshold set may include multiple time value thresholds.
[0072] For each time value threshold, each output data can be cut according to the time value threshold to obtain a plurality of texture data, and the time values of the texture data corresponding to the same time value threshold are the same.
[0073] For example, when the time value threshold is 1 beat, each output data may be cut to obtain texture data whose time value is all 1 beat.
[0074] Each texture data can be cleaned.
[0075] It can be seen from the above technical solution that this embodiment provides an optional way to obtain multiple texture data. Through the above method, various types of data can be input into REMI, and the output data of REMI can be processed to obtain multiple texture data, thereby improving the randomness and diversity of the texture data.
[0076] In some embodiments of the present application, the process of obtaining the reference texture of the target real music in step S1 is described in detail, and the steps are as follows:
[0077] S12, determining the target music type to be generated.
[0078] Specifically, the target music type to be generated can be determined in response to the user's creative needs.
[0079] The target music type is determined by multi-dimensional information.
[0080] S13, selecting a real music piece that matches the target music piece type from a plurality of real music pieces as a target real music piece.
[0081] Specifically, a real song matching the type of the target music piece may be selected from a plurality of authorized real music pieces as the target real music piece.
[0082] S14, using the texture of the target real music as a reference texture.
[0083] Specifically, the texture of the target real music piece may be obtained and used as the reference texture.
[0084] It can be seen from the above technical solution that this embodiment provides an optional method for obtaining reference textures. Through the above method, reference textures that meet the creative needs can be extracted to better complete the texture generation.
[0085] In some embodiments of the present application, the music type is determined by multi-dimensional information, and the same dimensional information may exist in different music types.
[0086] The music type can be composed of any one or more dimensional information of music style, music mood, music segment type and bpm.
[0087] The target music type to be generated may include any one or more dimensional information of target music style, target music emotion, target music segment type and target bpm.
[0088] In some embodiments of the present application, the process of step S2, analyzing the texture structure for each texture structure of the reference texture, updating the texture data with the highest matching with the texture structure among the texture data, and generating the target texture corresponding to the texture structure is described in detail, and the steps are as follows:
[0089] S20. Determine a first music attribute of the texture structure and a second music attribute of each texture data.
[0090] Specifically, the musical attribute of the texture structure can be taken as the first musical attribute;
[0091] The music attribute of each texture data may be used as the second music attribute, and each texture data has a unique corresponding second music attribute.
[0092] The first music attribute can be used to characterize the first parameter value of the corresponding texture structure in each dimension.
[0093] Each second music attribute can be used to characterize the second parameter value of the corresponding texture data in each dimension.
[0094] S21. Selecting a target music attribute having the highest similarity to the first music attribute from each second music attribute, wherein the texture data corresponding to the target music attribute is the texture data having the highest matching degree with the texture structure.
[0095] Specifically, the similarity between the first music attribute and each second music attribute may be calculated; and the texture data with the highest corresponding similarity may be selected from each texture data as the texture data with the highest matching with the texture structure.
[0096] S22: Based on the deductive method of the texture structure, the texture data having the highest matching with the texture structure is updated to form a target texture corresponding to the texture structure.
[0097] Specifically, machine algorithms can be used to learn the deduction method of the texture structure, and the deduction method of the texture data with the highest matching can be adjusted to obtain the target texture corresponding to the texture structure after the adjustment.
[0098] It can be seen from the above technical solution that this embodiment provides an optional method for generating a target texture based on the texture structure. Through the above method, the similarity between the musical attributes of the texture structure and the musical attributes of the texture data can be integrated to identify the texture data that best matches the texture structure, thereby improving the reliability of the searched texture data.
[0099] In some embodiments of the present application, the process of determining the first music attribute of the texture structure in step S20 is described in detail, and the steps are as follows:
[0100] S300, determining a first duration, a first density, a first stability, a first rhythm, a first paragraph trend, a first pitch width and a first pitch range of the texture structure.
[0101] Specifically, the first duration, the first density, the first stability, the first rhythm, the first paragraph trend, the first pitch width and the first range of the pitch may constitute the first music attribute;
[0102] The time value of the texture structure can be used as the first time value.
[0103] The first time value can be used to represent the time value of the corresponding texture structure;
[0104] The duration value varies with the target real music. Generally speaking, the duration value can be 0.0, 1.0, 2.0, 3.0 and 4.0.
[0105] The density of the notes in the texture structure can be calculated, and the density of the texture structure can be used as the first density;
[0106] The density can be calculated according to the following expression:
[0107]
[0108] Among them, x1 represents density; ClickNum represents the number of times the note appears, and dur represents duration.
[0109] The first density may be used to indicate the density of musical instruments playing the corresponding texture structure.
[0110] The stability of the texture structure may be calculated and used as the first stability.
[0111] The stability can be calculated according to the following expression:
[0112]
[0113] Among them, x2 represents stability; ColumnNum represents the number of notes belonging to the column structure; there are more than three notes played simultaneously in the column structure.
[0114] The first stability is used to represent the proportion of the columnar structure in the corresponding texture structure.
[0115] The rhythmicity of the texture structure can be determined and taken as the first rhythmicity.
[0116] Rhythm represents the special rhythmic form formed by the arrangement of the timing / density of the notes struck by the texture structure. The arrangement and combination are composed of uppercase and lowercase letters, and Arabic numerals. There are 81 rhythmic combinations for each beat (including blank cases), and the number of texture rhythmic combinations in a measure is 81 to the fourth power (about 43 million).
[0117] The first rhythmicity can be used to represent the rhythmic form of the corresponding texture structure.
[0118] The trend of the texture structure can be taken as the first trend.
[0119] The trend can be divided into three types: flat, up and down.
[0120] The first trend can be used to represent the melody change of the corresponding texture structure.
[0121] The tone width of the texture structure can be used as the first tone width.
[0122] The value range of the pitch width can be [1, B8-C2].
[0123] The first tone width can be used to indicate the number of semitones between the lowest pitch and the highest pitch of the corresponding texture structure.
[0124] The range of the texture structure can be taken as the first range of the sound.
[0125] The value range of the range can be [1, 64].
[0126] The range can be calculated by the following expression:
[0127]
[0128] Among them, x3 represents the range of the sound domain; MinDomain is the lowest pitch; MaxDomain is the highest pitch.
[0129] The first range of pitch can be used to represent the pitch interval of the corresponding texture structure.
[0130] It can be seen from the above technical solution that this embodiment provides an optional method for calculating the first music attribute. Through the above method, the texture structure and texture data can be evaluated for similarity based on the duration, density, stability, rhythm, trend, pitch width and range. Through multi-dimensional evaluation, the matching degree of the texture structure and texture data can be improved.
[0131] In addition, the second duration, second density, second stability, second rhythm, second paragraph trend, second tone width and second range of each texture data can also be calculated.
[0132] The second duration, the second density, the second stability, the second rhythm, the second paragraph trend, the second pitch width and the second range of the second musical range may constitute the second musical attribute;
[0133] The second density can be used to indicate the density of musical instruments playing the corresponding texture data; the second stability can be used to indicate the proportion of columnar structures in the corresponding texture data; the second rhythm can be used to indicate the rhythmic form of the corresponding texture data; the second trend can be used to indicate the melody changes of the corresponding texture data; the second tone width can be used to indicate the number of semitones between the lowest pitch and the highest pitch of the corresponding texture data; and the second range can be used to indicate the range interval of the corresponding texture data.
[0134] In some embodiments of the present application, the process of updating the texture data with the highest matching with the texture structure based on the deductive method of the texture structure in step S22 to form a target texture corresponding to the texture structure is described in detail, and the steps are as follows:
[0135] S220, based on the musical note interpretation mode of the texture structure, adjusting the musical note interpretation mode of the texture data that best matches the texture structure, and obtaining a target texture corresponding to the texture structure after the adjustment.
[0136] Specifically, machine algorithms can be used to analyze and learn the interpretation method of texture structure, generate random seeds, and input the random seeds into the machine algorithms. Using machine algorithms, based on the interpretation method of texture structure and random seeds, the interpretation method of notes of matching texture data can be adjusted, and the target texture can be obtained after the adjustment.
[0137] It can be seen from the above technical solution that this embodiment provides an optional way to update the texture data that is most compatible with the texture structure based on the interpretation method of the texture structure. Through the above method, the note setting method of the texture structure can be combined to update and adjust the note combination method of the texture data in a targeted manner, thereby further improving the harmony and layering of the target texture generated by this application.
[0138] Next, we will combine Figure 2 The music generating device provided in the present application is introduced in detail. The music generating device provided below can be compared with the music generating method provided above.
[0139] See also Figure 2 It can be found that the music generation device may include:
[0140] An acquisition module 10 is used to acquire a chord sequence, a reference texture of a target real music, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music, and the reference texture is composed of a plurality of texture structures;
[0141] An analysis module 20 is used to analyze each texture structure of the reference texture, update the texture data with the highest matching with the texture structure among the texture data, and generate a target texture corresponding to the texture structure;
[0142] A combining module 30, for combining various target textures to generate a music texture for performing the chord sequence;
[0143] The generating module 40 is used to generate music based on the music texture and the chord sequence.
[0144] Furthermore, the acquisition module may include:
[0145] An output data acquisition unit, used for inputting text prompt parameters, midi data marked with text prompt parameters or random seeds into the REMI multiple times to obtain multiple output data;
[0146] The texture data acquisition unit is used to cut each output data based on a preset time value threshold set to obtain multiple texture data.
[0147] Furthermore, the acquisition module may also include:
[0148] A target music type acquisition unit, used to determine the target music type to be generated;
[0149] a target real music selection unit, used for selecting a real music that matches the target music type from a plurality of real music pieces as the target real music piece;
[0150] The reference texture acquisition unit is used to use the texture of the target real music as the reference texture.
[0151] The target music type is composed of any one or more of a target music style, a target music mood, a target music segment type and a target bpm.
[0152] Furthermore, the analysis module may include:
[0153] A music attribute determination unit, used for determining a first music attribute of the texture structure and a second music attribute of each texture data;
[0154] A music attribute matching unit, configured to select a target music attribute having the highest similarity with the first music attribute from each second music attribute, wherein the texture data corresponding to the target music attribute is the texture data having the highest matching with the texture structure;
[0155] The target texture generation unit is used to update the texture data that has the highest matching with the texture structure based on the deduction method of the texture structure to form a target texture corresponding to the texture structure.
[0156] Further, the music attribute determination unit may include:
[0157] A value determination component, used to determine a first duration, a first density, a first stability, a first rhythm, a first paragraph trend, a first pitch width and a first range of the texture structure;
[0158] Among them, the first duration, the first density, the first stability, the first rhythm, the first paragraph trend, the first pitch width and the first range of sound constitute the first music attribute.
[0159] Further, the target texture generation unit may include:
[0160] The texture data adjustment component is used to adjust the note interpretation mode of the texture data that best matches the texture structure based on the note interpretation mode of the texture structure, and obtain the target texture corresponding to the texture structure after adjustment.
[0161] The music generation device provided in the embodiment of the present application can be applied to music generation equipment, such as PC terminals, cloud platforms, servers and server clusters. Figure 3 The hardware structure diagram of the music generation device is shown in FIG. Figure 3 ,The hardware structure of the music generation device may include: at least one processor 1, at least one communication interface 2, at least one memory 3 and at least one communication bus 4;
[0162] In the embodiment of the present application, the number of the processor 1, the communication interface 2, the memory 3, and the communication bus 4 is at least one, and the processor 1, the communication interface 2, and the memory 3 communicate with each other through the communication bus 4;
[0163] The processor 1 may be a central processing unit CPU, or an application-specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention, etc.;
[0164] The memory 3 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), etc., such as at least one disk memory;
[0165] The memory stores a program, and the processor can call the program stored in the memory, wherein the program is used to:
[0166] Acquire a chord sequence, a reference texture of a target real music piece, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music piece, and the reference texture is composed of a plurality of texture structures;
[0167] For each texture structure of the reference texture, the texture structure is analyzed, and the texture data with the highest matching with the texture structure among the texture data is updated to generate a target texture corresponding to the texture structure;
[0168] Combining the target textures to generate a music texture for performing the chord sequence;
[0169] A music piece is generated based on the music piece texture and the chord sequence.
[0170] Optionally, the detailed functions and extended functions of the program may refer to the above description.
[0171] The embodiment of the present application further provides a readable storage medium, which may store a program suitable for execution by a processor, wherein the program is used to:
[0172] Acquire a chord sequence, a reference texture of a target real music piece, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music piece, and the reference texture is composed of a plurality of texture structures;
[0173] For each texture structure of the reference texture, the texture structure is analyzed, and the texture data with the highest matching with the texture structure among the texture data is updated to generate a target texture corresponding to the texture structure;
[0174] Combining the target textures to generate a music texture for performing the chord sequence;
[0175] A music piece is generated based on the music piece texture and the chord sequence.
[0176] Optionally, the detailed functions and extended functions of the program may refer to the above description.
[0177] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.
[0178] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0179] The above description of the disclosed embodiments enables professionals and technicians in the field to implement or use the present application. Various modifications to these embodiments will be apparent to professionals and technicians in the field, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. The various embodiments of the present application can be combined with each other. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A music generation method, characterized in that: include: Acquire a chord sequence, a reference texture of a target real music piece, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music piece, and the reference texture is composed of a plurality of texture structures; For each texture structure of the reference texture, the texture structure is analyzed, and the texture data with the highest matching with the texture structure among the texture data is updated to generate a target texture corresponding to the texture structure; Combining the target textures to generate a music texture for performing the chord sequence; A music piece is generated based on the music piece texture and the chord sequence.
2. The music generation method according to claim 1, characterized in that: Obtain multiple texture data generated based on the deep model REMI, including: Inputting text prompt parameters, midi data marked with text prompt parameters or random seeds into the REMI multiple times to obtain multiple output data; Each output data is cut based on a preset time value threshold set to obtain multiple texture data.
3. The music generation method according to claim 1, characterized in that: Obtain the reference texture of the target real music, including: Determine the target music type to be generated; Selecting a real music piece that matches the target music piece type from a plurality of real music pieces as a target real music piece; The texture of the target real music piece is used as a reference texture.
4. The music generation method according to claim 3, characterized in that: The target music type is composed of any one or more of a target music style, a target music mood, a target music segment type and a target bpm.
5. The music generation method according to claim 1, characterized in that: The analyzing the texture structure, updating the texture data with the highest matching with the texture structure among the texture data, and generating the target texture corresponding to the texture structure, comprises: Determining a first music attribute of the texture structure and a second music attribute of each texture data; Selecting a target music attribute having the highest similarity to the first music attribute from each second music attribute, wherein the texture data corresponding to the target music attribute is the texture data having the highest matching degree with the texture structure; Based on the deductive method of the texture structure, the texture data with the highest matching to the texture structure is updated to form a target texture corresponding to the texture structure.
6. The music generation method according to claim 5, characterized in that: Determining a first musical attribute of the texture structure includes: Determining a first duration, a first density, a first stability, a first rhythm, a first paragraph trend, a first pitch width and a first range of the texture structure; Among them, the first duration, the first density, the first stability, the first rhythm, the first paragraph trend, the first pitch width and the first range of sound constitute the first music attribute.
7. The music generation method according to claim 5, characterized in that: The deductive method based on the texture structure is to update the texture data that has the highest matching with the texture structure to form a target texture corresponding to the texture structure, including: Based on the musical note interpretation mode of the texture structure, the musical note interpretation mode of the texture data that has the highest matching with the texture structure is adjusted, and after the adjustment, a target texture corresponding to the texture structure is obtained.
8. A music texture generating device, characterized in that: include: An acquisition module, used to acquire a chord sequence, a reference texture of a target real music, and a plurality of texture data generated based on a deep model REMI, wherein the reference texture is used to characterize a chord interpretation method of the target real music, and the reference texture is composed of a plurality of texture structures; An analysis module, for analyzing each texture structure of the reference texture, updating the texture data with the highest matching to the texture structure among the texture data, and generating a target texture corresponding to the texture structure; A combination module, used for combining various target textures to generate a music texture for performing the chord sequence; A generation module is used to generate music based on the music texture and the chord sequence.
9. A music texture generating device, characterized in that: including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the music generation method as described in any one of claims 1-7.
10. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, each step of the music generation method as described in any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Music texture generation method and device, electronic equipment and storage medium
CN115035884A
Accompaniment synthesis method and system based on texture interval relation analysis inverse structure
CN115862573A
Music accompaniment generation method and device, equipment, storage medium and program product
CN117765902A
Textile combination method and device, computer equipment and storage medium
CN118397987A
System and methods for automatically generating a muscial composition having audibly correct form
GB202104696D0
