You no longer have to break the bank for quality original music./ShutterStock
Don't miss the previous guides in the series
5 steps to building an AI chatbot that will upgrade your customer service system
How to create an AI avatar for business marketing or for any other purpose
How to create stunning videos with AI
First of all why?
Why would we want to create music with AI when there are so many talented musicians out there?
Well, like many other fields, the main barriers are time and budget and AI helps to break them with peak efficiency.
Without going into the moral issue, creating music using artificial intelligence can be very helpful for businesses, who may be less knowledgeable about music and don't have the budget to invest in musical content or buy copyrights.
In addition, it can also help the musicians themselves, since it optimizes and shortens processes, and enables personal expression and creativity in an extremely fast period of time.
The fields and possibilities in which artificial intelligence can help us create songs and music are almost unlimited, starting with writing words and notes for songs, "playing" melodies with various instruments, editing songs, creating cover versions of existing works, training a human voice duplicator and more.
In the article before you we will touch on each of these components, we will talk about the recommended tools for doing this work, and we will also show you step by step how to create music from scratch, which you can use for marketing purposes or for any other purpose.
Before diving into the technical details, it is important to understand what the main steps in the process are:
Brainstorming and inspiration
Writing lyrics to a song
Writing notes, basic melody and chords
Deciding on the song's structure, style and dynamics
Deciding on the musical instruments that will participate in the creation and writing of their roles
The creation of poetry and music
Connecting all the elements together
Polishing and sound editing
A natural solution
This is how you will improve performance and pleasure in bed - with an exclusive sale
in collaboration with "Gabra"
Brainstorming and inspiration through language models
Do you lack ideas for song topics?
Do you have an idea for a song about some topic but you want to brainstorm about it?
You will be able to use different language models in a variety of fields in the context of songwriting and music.
The paid versions of CHATGPT and Claude.ai which works wonderfully in Hebrew will give excellent results, but the free version of COPILOT will also do a good job.
If you have CHATGPT Plus you can find a variety of GPTs for music at this link.
To get started, just start a new conversation with the language model and write to it what you would like it to do for you.
You can use the following prompt structure:
Background
- general background about you and the task, for example a song for a diaper commercial, a waiting tune for a customer service center, etc.
Purpose
- What is the purpose you would like to achieve in this correspondence, for example getting ideas or writing a poem.
The role of the model
- for example a poet, composer or singer.
You can even ask him to act like a certain artist you like (note that if you ask the model to act or write like a specific artist for a commercial purpose you may run into copyright issues).
The product you want to receive
- for example lyrics to a song, notes to a song, chords, a chorus, a transition section, etc.
As mentioned, when working with language models, it is always recommended to specify and specify the instruction as much as possible according to the need.
Here is an example prompt:
For the 34th birthday of a friend named Yohai who really likes Borax (specify everything that is relevant) and works as a lawyer, I would like ideas for a funny song that I can sing to him on his birthday. Act like a comedian and write me 7 funny song titles.
After receiving the product From the initial model, you can continue the correspondence until you reach a product that you are satisfied with.
Writing lyrics to a song using language models
Besides inspiration and ideas, the language models can be used to write lyrics to the lyrics of your song.
For this, we will open a new conversation or ask the model to write us the song we want.
To get as detailed and rich a result as possible, be sure to include in the instruction everything that is relevant from the following details:
Musical genre
- for example pop, country, etc.
The main content theme
- love, friendship, protest, etc.
The length of the song
- short/medium/long or really specify a specific time.
Writing style
- poetic/ narrative/ metaphorical etc.
Emotion and atmosphere
- sadness, joy, hope, etc.
Musical characteristics
- eg specific rhythm or melody, stanza structure and line length.
Verbal characteristics
- for example preferred images and verbs, use of irony and humor or figurative language.
Cultural context
- the identity of the artist and the listener, and a story, event or characters that the song is about.
Example prompt:
I have a housewares store.
For an advertisement I'm creating, act like a rap artist and write me a short, humorous song with 3 stanzas and a hip-hop style chorus about cutlery and kitchen utensils, write from the point of view of a small fork singing to his lover's ladle.
Continue the correspondence until you receive the desired product.
If the song is in Hebrew , and you want to create artificial vocals for it, ask the model to add a score to the final song. You can score text easily with an automatic scorer. This will be very useful later when we get to the actual music creation stage.
Writing notes and chords using language models
We can also use the language models to write and compose notes or chords for songs.
We will ask the model to write us notes or chords for a certain musical instrument and we will send him the song.
Sample prompt:
I'm attaching you lyrics to a pop style song I wrote.
I want to match him with a rhythmic and fun accompaniment.
Act like a pop artist and write me chords in the F minor piano scale to accompany the song.
The lyrics of the song are: (here we will enter the song).
You can of course combine all of these processes or some of them together, and get a skeleton of a song, which you can use in the next stages of your creation.
After you have finished writing the words and notes, you can move on to the fun part - creating the actual music.
For this you can use the free SUNO or Loudly.
SUNO - creating music in different styles in full length and completely free
The first tool we will recommend is SUNO, a simple and convenient tool with excellent performance and good support in Hebrew, which is suitable for a wide range of operations and needs.
You can use it to create a melody for songs you wrote or created using the language models.
Advantages -
easy to use, good performance, allows you to create long sections, supports Hebrew and includes a free version.
Disadvantages -
there is no control over musical parameters beyond the style and it is not possible to enter musical pieces as a reference, in the free version there are no commercial usage rights.
Monthly price -
free / $10 / $30.
In the free version, you can only create five times per day, and without commercial usage rights.
Purchasing a subscription will give you access to advanced features and commercial usage rights.
So how do you make music with SUNO?
Once you have created an account and connected to the site, follow these steps:
Click on the
Create
button .
Enter a prompt
(also available in Hebrew) with the description of the song you want.
Be sure to include the song's theme and musical style in the description.
Click on
Create
again and wait for the result.
Create music according to preset parameters with SUNO
With the method we described above, you will not have much control over the words or the final product beyond the general theme and style, so I would recommend switching to an advanced creation mode that breaks down the process into several steps:
Click on
Custom Mode
above the prompt box, enter text and choose the style you want.
In the
Lyrics
box you can enter a text of your choice (also possible in Hebrew).
For an optimal result, it is recommended to enter in the area of 50 words at a time.
Alternatively, you can leave this box empty to receive only a melody, or select the
Generate Lyrics
option to generate random text (only in English).
As mentioned, if you are entering text in Hebrew, it is recommended to enter dotted text.
For the best result, you should indicate to the instrument where the verse begins and where the chorus begins, by adding the words [
Verse
] and [
Chorus
] inside square brackets in the appropriate places.
After you have finished entering the text, go down to the
Style Of Music
box and enter in small letters in English one musical style or a fusion of styles for example: pop, hip hop, electro pop, or choose
Use Random Style
to let the model choose a random style for you.
In the
Title
box , enter the title of the song.
Click the
Create
button and let the magic happen.
After a few seconds, you will receive two pieces of music for each piece, which you can download as an audio-only file or a video that includes the lyrics and audio together, by clicking on the three dots icon and choosing your preferred option.
The length of the segments will be determined by the number of words, but don't worry, if you liked one of the segments you can also continue it by clicking on the three dots and then on the Continue From This Clip button. Now you can re-enter the various parameters and create a continuation of your song
. To maintain consistency, it is recommended to enter the same musical style. Now you will again get two options for the continuation of your song. You can repeat the process again and again to create full-length songs. At the end of the creation, you can click on the three dots and then on the Get Whole Song button, which will allow You can download the full song you created to your computer.
Custom tunes with Loudly
Another excellent tool that allows you to create melodies with control over various parameters and without having to worry about copyrights is Loudly.
Advantages -
simple to use, control over many parameters, favorable terms of use and copyright
Disadvantages -
does not allow the creation of poetry, and the variety of styles is limited.
Monthly price -
free / $15 / $40 (60% discount on the annual route)
In the free version, you can create segments of 30 seconds in length, up to 25 times a month (three segments per creation).
You can download one segment per month, and use a single account on a social network.
Upon entering Loudly, 3 options will appear in front of you:
Generate Music With AI
that allows you to create customized music according to pre-selected parameters.
Text to music
feature
that allows you to create musical pieces using only a prompt.
Browse
feature
- an overview of existing pieces from a huge collection of copyright-free music, which can be filtered by styles and other parameters.
After you have created a Loudly account, go to the first option, choose a style from the list of different styles, then, in the toolbar on the left, add optional parameters to the creation:
Duration
- what is the length of the segment that will be created?
In the free plan you are limited to 30 seconds per creation.
Instruments
- What instruments will be used to create the piece?
The instruments available for selection will change depending on the musical style you selected at the beginning of the process.
Genre blend
- a sub-style that you can combine with the main style you have chosen.
Energy
- you can choose between 3 levels when referring to the level of intensity of the rhythm, melody and atmosphere of the song.
Structure
- choose the song structure from four pre-prepared templates.
Tempo
- Choose the tempo of the song by entering the BPM (number of beats per minute).
Key
- here you can choose a scale for the section.
Now all that remains is to press the Generate button at the bottom of the screen and the magic will happen by itself.
After a few seconds you will receive three sections according to the parameters you entered.
From here, you can click Studio to further edit the clips, download to your computer, or share with others.
Creating vocals and voice duplicating training with KITS.AI
So you created music and you want to add vocals to it?
With KITS.AI you can create singing parts with voices in different styles, and even create a duplicate that will sing in your voice.
Advantages -
compared to the tools of others in the field, the tool is very simple and easy to use, it is also possible to reproduce a voice in Hebrew.
Disadvantages -
the voice duplicating function is only available for paid subscribers, and the Hebrew support is only partial.
Monthly price -
free / $10 / $25 / $60
With all subscriptions you can use the voice library without copyrights, in the free version you cannot reproduce a voice, you will get 15 minutes for voice conversion and 300 characters for free narration per month.
After you have created an account in KITS, follow these steps:
Click on
Convert Vocals
in the center of the screen.
At the top of the screen, click on
Select a Voice
and choose from a variety of voices and singing styles.
Under
Vocal input
you can enter text (
Text to Speech
) in any of the supported languages, or upload an existing audio file (
Select voice file
), and receive singing in the voice of your choice.
When finished, click on
Convert
.
After a few minutes, your product will appear on the right side of the screen as an audio track, which you can download by clicking the
down arrow button
or share directly using
the link button
.
Download the file to the computer to combine it with a melody later.
For better singing results in Hebrew, you can easily train a voice model with a paid subscription to KITS.
After payment, follow these steps:
On the left side of the screen click on
Train
next to the plus button and select
Train a Voice
.
In the center of the screen, select
Start Training
.
Now you can
upload an audio file
from your computer or attach a link to YouTube.
For a better result, you can activate the buttons
Remove Instrumentals
to remove musical instruments from the file and
Clean up Vocals
to clean the audio, but it is more recommended to upload a clear audio file that initially does not contain a melody or background noise.
We will detail this later in the article.
It is also advisable to clean the audio file with a dedicated tool such as Adobe Podcast AI.
The maximum weight of the file you can upload is 150 megabytes, for best results it is recommended to use a section of 10 minutes or more.
After uploading a clean sound file, click
Continue
and choose a name for your voice double.
If you want, you can also upload a picture to represent him.
Click on
Continue
again and then on
Train
.
The training time can take from a few minutes to a few hours, depending on the length of the file you uploaded.
After completing the process, you can choose the voice model you trained: Repeat the steps for creating a voice from the previous section, when you click on the Select a Voice button, you can choose My Voices and then the voice model you trained to complete the process of creating the voice.
Separating existing clips into audio tracks with VocalRemover
Want to use a voice from an existing segment?
The VocalRemover tool that will allow you to isolate the vocals, music, and the various instruments from existing songs and tunes to easily create covers and remixes from them, use them as elements in new compositions, or train them to create a vocal duplicate (not before checking copyright issues, of course).
Advantages -
no need to register, full support in Hebrew, simple to use, completely free.
Disadvantages -
Does not create new content.
Price -
free for use once a day, 29 NIS per month when paying for a year in advance, or 47 NIS per month with a monthly payment.
The use of tools is easy and even in Hebrew.
Just follow these steps:
Click on "
Search in my files
".
Upload
a music file
from your computer or phone.
The tool will process the file and after a few minutes will give you the vocals and the various instruments separately.
By clicking the "
Save
" button, you can download one or more of the components separately and save them for later processing.
The site includes several other tools for processing music: changing tempo and scale, uniting components, cutting sections, karaoke and recording.
You can access all the tools using the menu in the upper left corner of the screen.
Splitting existing musical pieces into different audio tracks using Vocal Remover./Henry Stauber
How does it all connect?
Now that you know how to use artificial intelligence to create text, vocals, melodies and various components of your musical creation, comes the really fun part - combining and creating complex pieces with vocals and melody.
You can do this with any audio editing software, from simple and free tools like kapwing to professional editing programs like Logic.
When you're done, you can use the music you created for any purpose - whether it's background music for a commercial, a short video, or a full song that you'll play on the radio.
Remember that each tool you use has slightly different terms of use, and this will determine how you should proceed in terms of copyright.
However, if you created a good piece and took care of the rights issue ahead of time, there is no reason why the song you created will not become the next hot hit.
Good luck and most importantly - enjoy!
Henry Stauber lectures on artificial intelligence, manages the AI Music Gen Israel community and implements intelligence technologies in companies and organizations.
More on the same topic:
AI
artificial intelligence
music
content creators
guides