People should be more accepting of new technology. However, we would understand why people would mind using AI in certain ways. Recently, a tweet about an AI-generated song that imitates Jay Z’s style emerged. Understandably, people were less than pleased. ChatGPT, the topic of our article, can’t be used for that yet because it can only generate text responses. However, it can still do a lot in terms of making music. We sat down with the chatbot and asked it to help us make a song. Here’s our analysis of its response.

What is ChatGPT?

ChatGPT is an AI-powered chatbot that Open AI released as an advanced chatbot. Currently, two versions are available, ChatGPT and GPT-4. ChatGPT runs on the language model GPT3.5, and GPT-4 uses the language model of the same name.

GPT stands for generative pre-trained transformer. All GPT versions are based on the transformer architecture. They are trained by consuming large text data.

ChatGPT came out of the blue in late 2022, and GPT-4 followed in March 2023. The latter is more advanced in regards to creativity, interpretability, and accuracy. Moreover, it can analyze images, not only text.

For the sake of this article, we used GPT-4 to see what the most advanced AI technology-according to Open AI, has to offer.

How Can You Use ChatGPT in Terms of Music?

 

Here’s what you can ask the chatbot to do:

• Write full lyrics on any topic, mood, or style.
• Generate chord progressions for any topic, mood, or style.
• Compose a melody for the lyrics and chord progression. We asked the bot to compose a melody, and it came up with melody lines written in the musical alphabet. We chose not to include an analysis since we couldn’t try the melody out in the office.
• Change the lyrics and chords to fit a specific music style or genre.
• Analyze a song.

SendFame AI Music Generator: If you need something better than ChatGPT

SendFame’s AI Music Generator is a groundbreaking tool designed to revolutionize how music is created. By harnessing the power of artificial intelligence, SendFame allows users to generate their own unique music compositions effortlessly. Whether you’re a seasoned musician or a beginner, SendFame provides an intuitive platform where you can input preferences, such as genre, mood, or instruments, and receive original compositions tailored to your specifications. This tool empowers artists and creators to explore new musical landscapes, providing inspiration and a starting point for further creative development. With SendFame, the process of composing music becomes more accessible, allowing anyone to craft high-quality, professional-sounding tracks with the click of a button.

In the following sections, we showcase the answers that GPT-4 provided when we prompted it to make music and what we think of the output from the viewpoint of a songwriter. If you want a conclusion of what we think about GPT-4’s musical abilities, skip to the conclusion section.

Writing Lyrics

The general prompt we used for this article is to write about “finding yourself at the age of 45.”

We specifically told the chatbot to write a song about the matter. The chatbot provided a full song lyric, with the bridge and outro included. But to make things short, we will only show the first two verses and the chorus.

LYRICS

I was lost in the maze of life, at the age of forty-five,

Treading water, feeling small, wondering when I’d truly arrive,

A thousand roads I’d walked before, but none led to my core,

I’d been living a borrowed dream, but I couldn’t do it anymore.

(Chorus)

At the age of forty-five, I found my heart’s desire,

A blazing sun within my soul, igniting my inner fire,

I’d been searching far and wide, but the answer was inside,

In the mirror of my soul, I found myself, alive and whole.

(Verse 2)

I shed the skin of expectations, the weight of the world released,

In the silence of my solitude, I discovered my inner peace,

One foot in front of the other, I started to embrace the climb,

A journey of a thousand miles, one step at a time.

The Analysis

Reading the lyrics, the structure is fine. They are correct, with no grammatical or lexical errors. However, as a lyricist, I am inclined to see them as a bit of a cliché. It might appeal to a certain audience, though, and I’ve seen worse lyrics before. I can imagine the lyrics becoming a pop or country song, with all due respect to both genres.

Poetically, the first verse and the chorus rely heavily on rhyming, save for the last sentence in the chorus, which creates a nice contrast. The second verse has the same rhyming scheme (AABB) as the first verse, with a slight change in the first two sentences, making it a half rhyme. I found the rhyming too on the nose, but that could be a personal opinion.

To sum up, the lyrics are okay, but I don’t believe they would appeal to poets or word enthusiasts.

Lyrics and a Chord Progression

This time, we asked GPT-4 to write a chord progression about finding yourself at 45. The answer started with a disclaimer that GPT-4 can’t provide music notation and that you have to use your musical abilities. Here’s what we got.

Verse 1:

Em G D A

Lost in the maze of life, I roamed

Em G D A

Years slipped away, and time swiftly flowed

C G D Em

At forty-five, I found my place

C G D A

A journey’s end, a warm embrace

Chorus:

G D Em C

I found myself at forty-five

G D Em C

A second chance, a new sunrise

G D Em C

My soul awakened, heart revived

G D Em C

I found my purpose at forty-five

The Analysis

Again, we only pasted the first verse and chorus.

Let’s start with the lyrics. They are quite different from the ones that GPT provided before, much simpler and less wordy. Note that we didn’t mention lyrics in the prompts, but they were provided anyway. The lyrics are still rhyme-y, more or less similar in terms of being clichés, and it could fit different styles.

Now to the more exciting part, the chord progression. The whole song spans five chords, which is more than enough. At first, I couldn’t tell the song’s key, and you will know why shortly. So, I asked, and the bot answered that it’s in E minor, and the chorus is in its relative key, G major. I had to double-check to confirm that the renowned bot had made a mistake. Notice that the chords in the verse are Em, G, D, and A. The A chord isn’t part of the G major scale, but you have to have basic knowledge of music theory to be able to tell.

I asked the chatbot because maybe there’s something I am missing. It replied that I was correct and that it had made a mistake, then reposted the lyrics with the chord progression, with Am this time.

After the correction, I can say that ChatGPT didn’t do so badly in generating chords. But it was a basic progression, so I took it further.

Changing the Song’s Style

To make things trickier for GPT-4, we asked it to change the lyrics to be more R&B and jazzy. Here’s how it responded:

Sure, here’s a revised version of the lyrics with a more R&B/jazzy feel:

Title: “Forty-Five Groove”

Verse 1:

Am7 D9 Gmaj7 Cmaj7

Drifting through life, a wandering soul

Am7 D9 Gmaj7 Cmaj7

Time took its toll, as the years rolled

Fmaj7 Bm7 E7 Am7

At forty-five, I found my groove

Fmaj7 Bm7 E7 A7

A journey’s end, in rhythm I moved

Chorus:

Dmaj7 A7 Bm7 F#m7

Rediscovered at forty-five

Dmaj7 A7 Bm7 F#m7

A new chapter, where dreams collide

Dmaj7 A7 Bm7 F#m7

My heart’s rhythm, my soul’s delight

Dmaj7 A7 Bm7 F#m7

I found my purpose, in music’s light

The Analysis

Note that we didn’t omit any part of the lyrics, meaning they changed the whole structure to be more jazzy.

The title, lyrics, and structure have all changed. I was impressed. Now, the chords are more jazzy, all in the 7th, and contain notes outside the scale. Maybe, some chords were unintentional, but I couldn’t tell.

Even the lyrics are befitting, using the right words and sentence structure. Again, I imagine that a human can write better— or worse.

Musical Analysis

Next, we decided to use ChatGPT for something different. We asked GPT-4 to analyze a song from a musical standpoint. For this, we chose “Hey Jude” by the Beatles. The results were more elaborate than we had imagined. They did lack accuracy, though.

At first, the chatbot introduced the song, its release year, and what made it iconic. Then, it analyzed the song musically, dividing it into numbered points.

• The first point was about the song structure, dividing it into timestamps, each labeled as verse, chorus, or outro. The division was inaccurate in some parts.
• The second was about the key and harmony, mentioning the song’s key in detail and how it contains chromaticism in the chord progression and modulation in the chorus, which is arguable. One could argue that the song doesn’t have a chorus.
• Thirdly, the bot elaborated on the melody, mentioning how it contains large leaps stepwise. It added the contrast between the verse and chorus, as one has a descending melody and the other has an ascending melody. It could have been true if the song had a different structure.
• The fourth point is about rhythm, which was more correct than the previous parts. It mentions that the song is in 4/4 and has a syncopated rhythm, which isn’t wrong.
• The fifth part mentions the instrumentation and who plays what. The sixth mentions the outro as the song’s most iconic aspect and analyzes it from a critic-like point of view.

Overall, this would have been nice if it were accurate. It has great potential, but after all, the bot derived the answer from data and didn’t listen to the song. It left a lot to be desired.

Conclusion

GPT-4 can help musicians make music. But help is all it can do. Without my knowledge of music theory, I would have used that A chord in the Em chord progression. A musician would know by ear or knowledge when something was off. However, it can help a songwriter stuck on what to add next. It’s debatable whether using a chatbot to make music is cheating. In an ideal world, people would use their ears and expand their knowledge to achieve the most creative and unique result. Our world is far from ideal. We predict that because some people in the music industry only care about making money, they will overuse it to mass-produce mediocre songs.

With all being said, there’s more to music than melody and chords. Producers, vocalists, and instrumentalists can do much to add a human touch. That’s why we don’t see GPT-4 as a threat to the music industry yet. We never know how it will progress in the future, though. Other AI systems are a different story, but that’s off-topic.

Are you a musician or manager looking to build an app for your band? Forget about the costs of custom development and the wait time. With no code, you can create an awesome app today! Sign up for the free trial now.