Brief Summary
This video explores the science behind hit songs, aiming to uncover the "formula" that makes certain tracks successful. It examines various scientific studies and analyzes a vast collection of songs to identify the key elements that separate hits from misses. The video covers aspects like genre selection, beat creation using Spotify data, song structure based on psychological experiments, melody composition, and lyric writing, all while creating a song to test these findings.
- Hit songs often have simple melodies, lyrics, and ideas.
- Unpredictability and predictability are used to hook listeners.
- Repetitive patterns in melodies, song structure, and drum grooves enhance engagement.
- Hit songs evoke strong emotions through melody, beat, production, and lyrics.
The video introduces the concept of a scientific formula behind hit songs, suggesting that success isn't just about vibes or luck. The creator aims to find this formula by analyzing scientific studies and over a million songs from 1950 to the present, while also creating a song to illustrate the process. The goal is to understand what makes people repeatedly listen to certain songs and how to use this knowledge to craft hit songs.
The creator discusses the initial problem that sparked the investigation: an algorithm developed by Stanford researchers that could predict Billboard hits with 75% accuracy. Further research revealed another algorithm with 88% accuracy. The creator questions what patterns these algorithms are identifying in hit songs that differentiate them from others. To test the ability to predict hits, the creator quizzes themself on past popular songs, achieving only 36% accuracy, highlighting the difficulty in predicting hit songs compared to the algorithms' success. Studies confirm that people generally struggle to predict hit songs, with accuracy rates often below 50%.
The video transitions into the process of creating a song to test the hit-making formula, starting with genre selection. Research indicates that pop music is the most popular genre worldwide, while hip hop and R&B dominate the US market. Despite considering hip hop due to its US popularity, the creator decides to produce a pop song, aligning with their expertise and international appeal.
The focus shifts to creating the beat, utilizing a data set of Billboard Hot 100 hits from Kaggle. This data set includes Spotify metrics like danceability, energy, speech, acousticness, liveness, instrumentalness, and valence, which were also used in the 88% accuracy study. Spotify's definitions of these metrics are explained: danceability (tempo, beat strength, rhythm consistency), energy (loudness, timbre, entropy), speech (amount of talking vs. singing), valence (happiness or sadness), acousticness (rawness of the song), and popularity (number and recency of plays). The creator decides not to use popularity for beat creation, as it is more useful for predicting hits than making them. Averages from the data set are used to establish a starting point for the song's characteristics. The creator combines averages from their initial data set with those from another study to create a range of values. Reference songs are selected based on these ranges to guide the beat creation process, including songs for danceability, energy, acousticness and speech.
The video addresses song structure, referencing a study that analyzed chords in classic Billboard pop songs and found that musical pleasure arises from the right balance of uncertainty and surprise. The study suggests that listeners enjoy unexpected chords in simple progressions and predictable elements in complex progressions. An experiment involving rats and musical sounds is presented, revealing that rats maintain attention when presented with unpredictable note patterns, such as AAB ABC. This pattern is then related to song structure, with A representing verse, B representing chorus, and C representing bridge, forming a popular verse-verse-chorus-verse-chorus-bridge structure.
The video transitions to melody creation, citing a study that melodies are more important than lyrics in conveying emotional responses. Research indicates that melodies have become simpler over time, possibly due to increased complexity in other song elements like tempo, vocals, production quality, and timbre. The creator records various melody ideas that align with the study's findings, aiming for simplicity.
The focus shifts to lyric writing, noting that lyrics, like melodies, have become simpler and more angry over time. Studies show that sad lyrics intensify sad songs, while lyrics in happy songs don't necessarily increase happiness. Modern lyrics use more personal pronouns and fewer words related to positive emotions. The creator brainstorms simple, slightly angry words with personal pronouns, aiming to fit the Spotify song framework. A study on lyrical repetition is mentioned, stating that repetitive lyrics increase processing fluency and drive market success. Another study indicates that songs with higher rhyme saturation and basic lyric readability achieve higher chart positions and resonate more deeply with listeners. The creator focuses on writing simple, rhyming, and repetitive lyrics with a slightly negative valence.
The creator records vocals, produces the track, and adds background vocals to polish the song. A snippet of the final song, showcasing the first verse and chorus, is played.
The video concludes by reflecting on whether the hit song formula was found and successfully implemented. The key takeaways are that hit songs tend to be simple, using simple melodies, lyrics, and ideas. They incorporate unpredictability and predictability to engage listeners, and they use repetitive patterns in melodies, song structure, and drum grooves to enhance engagement. Most importantly, hit songs evoke strong emotions through their melody, beat, production quality, and lyrics. The video also acknowledges that factors not covered by the science, such as music marketing, social media, cultural trends, and playlist placements, significantly impact a song's success. Despite algorithms predicting hits with high accuracy, passion and hard work can lead to unexpected success.
Summarize any YouTube video in seconds
Get summaries like this one for any video, in your language, with the free Summ app.
Transcript
Full transcript of “The Scienitific Way to Make Hit Songs”. Click a timestamp to open the video at that moment. Need another video? Use the YouTube transcript generator.
0:00 why do some songs blow up While others fade faster than last year's Tik Tok Trends well it turns out there's actual science behind it yeah science not Vibes not luck but a formula that makes people hit repeat over and over again to find that formula I'm diving into a mountain of scientific studies and analyzing over a million songs from 1950 to today as I go I'll be piecing together what separates a hit song from the rest all while making a song of my own let's see what we discover and how you can use it to craft your own hit songs and to understand why I chose this topic let's take it back to where this problem all started so I was just scrolling online but I came across this article that kind of blew my mind apparently a group of researchers from Stanford built an
0:47 algorithm that could correctly predict Billboard hits 75% of the time what that seems pretty high right I feel like most of the time the music I like and listen to is way different than what trending on the Billboard Hot 100 hits we do a little bit more research so I did some more research but what I found surprised me even more so I found another research study that kind of corroborates what I found and the first research study this algorithm can predict songs with 88% accuracy that's 133% higher than the last one if this is true and they could predict it with this high of accuracy what are they looking at what is the patterns that they're finding between these hit songs that the other songs don't have and that very question led me
1:34 down the rabbit hole I was about to jump into today before going any further let me just make sure that 88% is really that good as good as I think it is so in order to do that I'm just going to quiz myself using chat gbt I'm going to ask it to give me a list of 80 popular songs over the last 50 years and I will guess if they got on the Billboard Hot 100 or not little did I know how hard this quiz would be pretty good if I to guess how what percentage I got I I feel like I got 80 70 70 80% we're going to find out turns out I was wrong we're going how how much is it what no wait wait what 36 that's not right is that right yes it was correct turns out
2:21 I only got 36 out of 80 right I even recounted every vote but that made me think if a professional music producer like self did that badly how would the general population fair so as it turns out pretty much the same as me this study I found was analyzing the entertainment industry and movies specifically people were trying to guess if whether they watched a movie if it was going to be a success or not they got it right around 50% of the time I also found this study by Robert prey that just generally said people are bad at predicting hits then I found this study which was using functional MRIs to determine what was going to be hit songs and you guess it it got it right less than 50% of the time turns out we are not good at predicting what's going to be successful or what's going to be a hit but algorithms apparently are there
3:09 has to be something more that this algorithm is looking at that I'm just not understanding right now and I was determined to find out what that was the best way I can think of to dissect what's going on behind this algorithm is to one look at some more scientific studies to see what the research says and two build the song along the way to see if I can make it sound like some of these Billboard hits and if we're going to start off writing a song the first thing I want to know is what genre I should write so I hopped on my computer to see what the scientific research says according to this article from headphones addict pop is the most popular music genre worldwide that's I mean that's what I was thinking pop is just short for popular music so that
3:54 makes sense whereas hip hop and R&B dominates the US market I mean I'm in the us so maybe that means I make a hip-hop genre because it's the most popular giving it the best shot of charting but I am a pop music producer and I am very International I'm not I'm not International but I do think I am going to make a pop song instead so even though I had the genre selected I still had to determine what I should actually use for my beat so I did some research and I actually found a really cool data set that had a bunch of Billboard Hot 100 hits off of kaggle I put them into into a Google sheet and it gave some really cool information as you can see here all of these songs that you see on the left are ones that have charted I had no idea that Spotify was taking this much data when you upload your songs but I guess
4:41 it is it tracks something called danceability Energy speech acoustic liveness instrumental veilance what and actually according to this 88% study if you look at the details you'll notice that it used these Spotify details and all of these values I just mentioned in that algorithm so maybe cracking this code and understanding what each of those numbers really mean is going to help us make our beat let me quickly explain what Spotify means by each of those different adjectives that they use in their data danceability refers to how easy it is to dance to the song apparently Spotify measures this by things like Tempo beat strength and the Rhythm consistency that it provides energy is how calm or exciting a track feels Spotify measures the general loudness the tamber of instruments also
5:28 the entropy or the amount that chaos is happening in the song I'm not a data scientist I'm just just reading this speech covers how much of the song is talking versus how much is singing Spotify has a bunch of podcasts so his podcast would be closer up to one rap songs would be more speechy for example than say a love song ballad that you sing the entire one I have it sorted by speechin too which is why this is all kind of collected together rather than the scatter plot we were seen before vilage shows how happy or how sad a song is and by looking at this chart it looks like all across the board there's are tons of happy and sad songs that make the Billboard Hot 100 this this is great this is this is going to help me for sure but as it turns out it only made things harder Spotify measures how happy or sad a song is by
6:14 things like the lyrical content major or minor chords that it uses and the type of instruments the song uses as well acoustic is how raw the song is for example the best way I can describe this is if you were performing this in a coffee shop with just an AC acoustic guitar and singing that would be very acoustic whereas if you have a lot of production quality and a lot of production sense that makes your acoustic go down and as you can see most of the songs are gearing towards more the side of higher production value you can see a high skew on the very low end of things and finally the one that I think might have helped these algorithms the most is called popularity popularity uses the number of plays and how recent those plays are to rank a song on how popular it is this is probably great for
6:59 predicting hits but it's not going to help us make our beat so I won't be using it if I do an average of all of these columns I'm going to have a really good starting point for where to put my song and yes this was a great starting point but this was only 600 songs there was way more Billboard hits out there that I could be pulling from so I decided to hop back into the research so I also found this study using a different data set on Billboard hit songs that did something similar and they put it into a nice chart that I can use and show you right now what I'm going to do is I'm going to use these averages put them with my averages and combine them into a range that I can use then I'm going to use that range and actually find some reference songs that I can listen to and get an idea of what 6 dance ability actually might mean in this case we'll have to make a song as happy or sad as hey Mama by David geta
7:46 or I don't give a by do a Lea in terms of energy we're shooting for something like say so by DOA cat or flowers by mile Cyrus for danceability we're looking at something like Star Boy by the weekend or Since You've Been Gone by Kelly clent for acoustic we're looking for something like thank you next by Ariana Grande or Take a Bow by Rihanna for a speech we're looking for something like Juice by lizo or run it by Chris Brown okay this is a lot of references though normally for references I don't choose more than three songs that I'm going to base my song off of this is like 12 songs but I do think it is going to help us make the beat so I'm back I spent the rest of the day working on the beat and kind of fleshing things out I also have happened to get a haircut now I'm ready to take some of these ideas I had and arrange
8:32 them into a song and kind of put them where the song structure needs to go like what what does science say about song structure itself should I have a verse should I start with A Chorus do I even need a bridge well let's find out from one study that analyzed 880,000 chords over 745 classic us billboard pop songs they found that musical pleasure comes from the right combination of uncertainty and surprise the study basically says two things when people are listening to simple chord progressions and an unexpected chord happens they tend to prefer that and also if they're listening to a complex core progression and something predictable happens or simple then they also enjoy that I also happened upon another study which I thought was really interesting in order to study expectation and psychology Dr David
9:19 hiron created an experiment that used musical sounds with rats he was determined to see how many sounds could be played that could scare a rat's attention into turning towards it started by playing one note an a the rats would turn their head then he played a a the rats turned their head both times but when he played a AA on the third a the rats stopped caring for that third sound the rats didn't turn their head they just weren't interested but something interesting happened when he changed it up instead of playing a AA Dr hon played a a b this time the rats had their attention the entire time all three times they turned their head he continued with this experiment and tried to see what the shortest number of letters that he could continually keep rat's attention for he ended up finding this pattern a a b a b c just like the
10:08 last study it seems like the unpredictability of these notes is what made them interesting but what's even crazier about this particular pattern is I think it's perfect for song structure if we replace a with verse B with chorus and C with Bridge we have a blueprint for what we can do with our song in this case it would be verse Verse Chorus Verse Chorus bridge that also happens to be one of the most popular ways of structuring songs that we have and with that piece of information under my belt I went back into my song and started arranging it according to what the study had told me and once I was happy with the arrangement of the song I was ready to move on to one of the most important parts of this hit the melody let's see what science has to say about writing a good Melody I found this study that specifically compares the
10:54 impact that lyrics have versus Melodies and it looks like that Melodies are actually more important than the lyrics in conveying emotional responses I've been preaching this for years so it's nice to be backed up I also found this study that showed me that over time Melodies have continued to get less complex and simpler this study also mentioned that the reason in simpler Melodies maybe because an increase in complexity in other factors of songs those factors could be faster tempos more vocals and more vocal layers higher quality of sound production and an increased overall complexity in tamber of the song itself I completely agree with that ever since the transition from the analog world to the digital all of those tools have become way more accessible than before and that's all well and good but a simple and non-complex Melody can't be all that I'm
11:40 shooting for right well let me hop into my beat and make something that I think would fit in this lens I went ahead and recorded a lot of different Melody ideas that fell in line with what the studies were telling me to do and with each part I added my song began to sound better and better but there was still one important aspect that I needed to nail if my song had any shot of success the lyrics the song is sounding a lot better but it still needs some lyrics I really enjoy writing lyrics so I'm excited to see what the science says about this according to this study just like songs lyrics Have Become simpler and less complex over time it also mentions that lyrics have become more angry over time interesting this study showed that sad lyrics can make sad songs even sadder
12:25 whereas lyrics on Happy songs don't necessarily make happier and in this study it shows that lyrics nowadays are using more personal pronouns than ever before like I and me it also noted that words related to positive emotions like love are being used less before I start writing my lyrics I'm going to brainstorm some words that fall in line with these studies that are simple that tend to be a little bit angrier that use some more personal pronouns I'm also going to try to use these lyrics to fit into that Spotify song framework that we talked about before obviously the lyrics and the lyrical content is a big reason if people will feel happy or sad with the song Let Me hop into the program and just come up with some ideas I started by creating a word list of potential words I wanted to include in my lyrics I'll use these words later as a pallet to pull from as I'm brainstorming
13:11 lyrical lines for this song I tried to keep the lyrics simple and following some of the studies that we' previously talked about and although I did feel pretty confident in this pallet of words they weren't going to mean anything unless I actually wrote some lyrics down okay so I had to take some time away but I'm back on day three with my brainstormed lyrics here but before I do that I want to hop back into the science and see if it can guide me in any way this study from the Journal of consumer psychology showed the power of lyrical repetition saying that repetitive lyrics in a song increase processing fluency and drive Market success I also found this study which showed me that songs with higher rhyme saturation and basic lyric readability not only reach higher chart positions but tend to resonate more deeply with listeners both of those are great studies for just pointing out
13:57 that I need to be simple with my lyrics but I also need to include a lot of rhyme on top of that I should be repeating things pretty frequently so with that in mind let's get writing I started off writing my lyrics by using the previous brainstorm words that I had come up with as a guy but this turned out to be far from Easy although I had the melody previously etched out I was really trying hard to fit in the right syllables to make the track as dancable as the references all while at the same time trying to make my lyrics simple repeatable and then have a slight negative veilance this ended taking me hours but if my song were to have any shot of success I needed to get it right I think I've got my lyrics finished up here so I'm going to record these vocals I'll do some producing on the track and then I'll add in some
14:42 background vocals or just polish it up so it sounds professional and then I'll show you the final result after all that it was time to see if the scientific studies I followed were going to make this song sound like the [Music] hits still got a r Je she left in my room now she just history and I'm all news we had a problem F love weeks apart she's calling me up
15:30 and I'm not sure if I'll pick up is it a Wast of time if you came to stay the night fall asleep with me right by By Your Side by your side tell me is it worth the RIS will it Hur to make it worse again should be turned away what we did with one last night and one more so there it is I'm only going to show the first verse and chorus if you want to see the full song go and like this video this gets 1,000 likes I'll post it on Soundcloud you can hear the
16:16 full version but now the question we've been waiting for was I able to find the formula that really separates the hits from the mediocre was I able to use it in the song I think kind of we definitely learned a lot along the way in what sign science has to say about hit songs it seems that the majority of billboard hit songs are simple using simple Melodies Simple lyrics and simple ideas they use unpredictability and predictability to hook the listeners in they use repeated patterns in Melodies song structure and drum groups to hook people in even more and maybe most importantly it seems that hit songs make you feel something more that your Melody the beat your production quality and lyrics convey that emotion the higher likelihood it has of success but even with all of those ideas there are a lot
17:02 more factors that the science didn't cover that I'm sure have a big impact on a song success including music marketing social media cultural Trends playlist placements and more an algorithm may be able to predict 88 out of the next 100 songs that will go on to be hit but if you keep working hard and you keep making music you're passionate about you can be part of the 12 that it never saw coming
