Third edition, ©November 2008, August 2011
This web page is the handout I use when I teach songwriting. In it, I describe one sequence of processes you might go through in order to write a song. I come as close as I can to the way I actually work, except that I break things into separate steps when I describe them. When I am actually working on a song, I overlap everything. I’ve found that when you’re learning, it help to work on one thing at a time.
When I write a song, I have a pretty good draft of the lyrics before I start to work on the melody, and this is the way I’m presenting the process in this guide. Note that lots of really good writers go about it the other way around. There is an appendix about tunestarting.
You walk into your workshop and grab a hammer and a saw and a few boards and start hammering and sawing away. In comes your significant other, asking, “Whatcha making, Honey?”. You grin your best devil-may-care grin and reply, “Dunno yet, but it’s shaping up nicely.”
Right. Of course you have to know what you’re doing, not just how to do it. You may be lucky here. The reason you are writing this song is because you just got a great song idea, or met somebody you want to express your tender feelings for, or got handed a songwriting assignment. Otherwise, you are going to have to spend a certain amount of time and energy deciding what you want to write. The best time to do this is NOT your songwriting time. It can be a procrastination mechanism that can suck up hours. Jot down ideas in your address book whenever they come to you, and use one of them now. Or start your songwriting period with a few minutes of free writing and use whatever comes up.
If you start right out trying to write rhyming lines, you are in great danger of getting so critical of the form that you can’t think about the content. So once you have picked a topic, write down everything you can think of that is relevant. Write for a good long time, so you will get lots of material. You may find free writing to be a good way to get words onto paper, or you might put your subject into the middle of a big sheet of paper, draw a circle around it, then put related things in other circles with lines going to the main topic or to each other. Some writers make lists of phrases describing the subject and lists of dramatic words that might fit in. It never hurts to do research, but it takes time. If you are writing about something you don’t know personally, talk to somebody who knows, or read a book about it, or surf the Internet. I am going to define research as step 0.5, because there is no way to do much of it during a songwriting class.
One clever idea is enough for a fairly successful song. A song hook is a turn of phrase – or, much less often, a snatch of melody – that gets everybody’s attention. It can be something as poetic as “The answer is blowin’ in the wind” or as un-poetic as “You flushed me from the bathroom of your heart.” Many popular songs are built around hooks. If ideas for hooks come to you, jot them down with your other songwriting ideas.
A really good song can be built around a hook. Nobody said you have to stop writing great lyrics (and great melody) just because you already came up with a zinger.
A good title really helps. Once people know the song, they will like it or not based on its merits and their tastes, but who’s going to bother to play a song named “I Feel Sad” or “Song #173” by an unknown writer? There is a Nashville convention that the title of the song must be in the chorus, which must mean that occasionally a good working title gets replaced by some phrase that doesn’t describe the song as well.
If you start with, or come up with, a good hook or title, write it down, circle it, underline it, then free write about it. DON'T STOP HERE! Or if you do, know that you're going to have to start again and generate a lot of stuff. One of the hardest things that will happen to you as a writer is you will almost finish a song, then get hit with a great hook that almost fits, and have to rewrite the whole thing.
Now you’ve written down much more than will ever fit into a song. Oh, you haven’t? Go back to Step One. Until you’ve had a lot of practice, you can’t think well about form and content at the same time.
Now you have enough stuff, as in more than you can use. It’s time for the first technical work. A good song is not just a whole lot of words that rhyme, it has a structure. The structure depends on what kind of song it is and what genre you are writing in. For instance, a song could tell a story, or describe a scene or a feeling, and it could be a ballad, a jazz song, a rock song, a rap . . .
Most of the kinds of songs we are dealing with consist of a few verses, possibly with a chorus sung after some or all of the verses, and maybe with a bridge to change things a little. Go through the stuff you wrote in Step One and select three or four particular things to say. Collect the stuff relevant to each of these points. Presto, verses! Summarize what the song says and call it a chorus. Something to say that’s a bit different but still want to get included? A bridge. What you have now resembles a very rough draft of either a short short story or an essay.
You could make an outline here, to help organize your material into sections. Or you could use a generic outline like this one:
A GENERIC SONG OUTLINE
Outlines are the safest things to borrow. Make an outline of one of your favorite songs, use it with a completely different topic, and a small part of your work is done. This is a stage more detailed than just using an overall structure, like verse-verse-bridge-verse or three-verses-and-a-chorus, since you are forcing yourself to look at what other writers are doing with the structure.
You will encounter lots of songs that don’t rhyme exactly, or at all. You aren’t going to find many that don’t have a rhythm to them, especially not in English. Western music is almost always organized into matching sets of lines, each line containing a related number of bars, and each bar containing the same number of beats. In order to be sung, a lyric must fit some pattern of beats and bars. Beats aren’t the same thing as syllables. You can patter several syllables into one beat, or stretch a syllable over several bars. What you can’t do is ignore the pattern of stressed and unstressed syllables that characterize normally spoken English.
That was a boring paragraph. Imagine trying to chant it. I am now going to rewrite it to have a chantable rhythm. Probably it will be even more boring, but I decided to do it just to prove that I could.
Lots of songs don’t rhyme at all
But every song has rhythm.
Melodies have beats and bars
And lyrics must fit in.
Syllables that don’t hit beats
Are perfectly all right,
But if you want to sound real strained
Just ignore the stress.
Prosody is the craft of making words expressive and easy to chant. By far the biggest part of it is setting up a structure of stressed and unstressed sounds, then choosing words so that stressed and unstressed syllables fall in the appropriate place. Structured poetry consists of sets of short patterns of stressed and unstressed sounds. These short patterns are called feet, and have been given special names which you can ignore. For instance, a foot consisting of soft–LOUD is an iamb, and a line of five iambs is iambic pentameter, and a set of a few thousand iambic pentameters is a Shakespeare play.
For contrast, a limerick is built of three-syllable feet, with optional additions or elisions at the beginning and end of each line. A poet would tend to analyze a limerick into amphibraches, soft-LOUD-soft feet, like this:
(There WAS a)(young GIRL from)(Van-COU-ver)
but written music tends to have the accent at the beginning of each bar, so somebody writing a tune for the limerick line would do this:
There | WAS a young |GIRL from Van - |COU –ver
treating the first word as a pick-up note before the first whole bar begins.
A melody in 4/4 or 2/4 time leads to lyrics with two-syllable feet. 6/8 melodies lead to three-syllable feet. Waltz time doesn’t correspond closely to spoken English, and leads to songs where some of the beats are split into two syllables.
(I was)|(DAN)-(cing)(with my)|(DAR)-(ling)(to the)|(TEN)-nes-see|WALTZ
Once you have the stresses sorted out, you can work on the finer points. There are short and long syllables – for example, the word “stretched” is technically one syllable but you wouldn’t use it in a passage where you are cramming a bunch of short notes together. Some vowels and ending consonants lend themselves well to long notes – that’s one of the reasons the moon turns up in so many songs.
And don’t forget the poetry! Words have associations as well as dictionary definitions. Use the word that gives the effect you are after, but maybe you’ll have to juggle with word order so it will fit the meter. And use your thesaurus (paper or software). You might find a synonym that scans better and means just as much.
You have a collection of lines that you can chant. That’s almost enough, but the people in your audience have brain wiring that lights up when two words rhyme. It makes them sit up and pay attention. It makes your song easier to remember. If you don’t want your song to be noticed or remembered, well, okay, you can skip this section.
Some very good writers deliberately avoid rhymes, or use vowel-only rhymes, because they feel that the jingling of perfect rhymes detracts from the experience of their message and melody. Some unambitious writers avoid rhyming because it is too much work. Using a mixture of exact and near-miss rhymes tends to sound as if you wanted a rhyme and couldn’t find one. This workshop concentrates on rhyming songs.
Look at the dull chantable paragraph again, with a couple of rhymes added:
Lots of songs don’t rhyme at all
But every song has rhythm.
Melodies have beats and bars
And lyrics must work with 'em.
Syllables that don’t hit beats
Don't make too bad a mess,
But if you want to sound real strained
Just ignore the stress.
A perfect rhyme requires that a stressed vowel sound and every sound after it match perfectly. Examples are be-me, beat-meet, beater-meter, teetering-metering. Rhymes of more than one syllable don’t have to be single words. Strictly speaking, “with ‘em” does not rhyme with “rhythm”, because there are two different “th” sounds in English and they don’t use the same one. You can get away with slightly imperfect rhymes by using clever words. Sometimes. Maybe not this time. Maybe I should replace “rhythm” with “meter” and work in a neater or some saltpeter.
As you did with scansion, you can move awkward words to a different position, or use a rhyming dictionary (paper or software) to see what options are available. Or go through the alphabet under your breath: “I need a rhyme for ‘oblige’ .. ige, bige, cige, dige, eige, fige . . .”. If you do this, don’t forget thige and chige and shige, and you might want to conside trige and swige and splige. It’s not a bad way to get a feeling for sounds, and it drives my spell checker bonkers. (You want frige with that?)
Not every line has to rhyme. Common rhyme schemes for a four –line verse are AABB, ABAB, ABCB, and AAAB. If there’s a really important word that you want to have at the end of a line where it can swell into a long note, and you can’t find a good rhyme for it, use a scheme where there are some unrhymed lines, and put it in one of those.
One final point about rhyme: people listening to a song come to expect a rhyme at a particular place, but they also try to guess which rhyming word you will use. The rhyme satisfies them, the word bores them. If you rhyme “moon” and “June” or “love” and “above”, they will feel that you chose the second word just to make the rhyme. One good fix is to put the common word first. In the Billy Rose example in the Bumper Stickers section, it would sound stupid to use the word “macaroon” and then rhyme it with “moon” or “June” or “spoon” or “soon”. But if you use it as the word that completes the rhyme, it will come as a surprise and let people enjoy the song more. (Of course, it's still a challenge to set up a context where "macaroon" doesn't sound idiotic. That was probably part of Rose's point.)
You have a set of lines which rhyme and scan and say what you want to say. That’s half the job. Now you need a melody to fit the words.
Ordinarily, when you are writing a melody, you spend some time thinking about the structure, but when you start with a set of lyrics, you know how many beats are going to be in a bar, how many bars are going to be in a line, and how many lines are going to be in a verse. You even know where the long and short notes are going to be. Now what?
You could begin by chanting the words while strumming chords. Sing mostly notes that fit the chords. Change chords when the feeling changes, or when you feel that you’ve played that chord long enough. Follow a chord structure that goes with the kind of song you’re after. Do this for a while and see if a tune emerges. Probably it won’t, but it’s worth trying.
Okay, that didn’t work. Try this: hum a scale a few times, then sing the first stressed note of the song to the first or fifth note of the scale. Pick a chunk two or four bars long and call it a phrase. Sing the last stressed note of the phrase against the first note of the scale. Assign the rest of the stressed syllables to the first, third, or fifth notes of the scale, then connect them using other scale notes that go up and down one or two steps, thusly. (In this case, the lines are short so a phrase is a whole line, but be aware that longer lines will probably break down into two or more musical phrases.)
Lots of songs don’t rhyme at all.
(5 Lots) of (5 songs) don’t (3 rhyme) at (1 all)
|5 6 5 4 | 3 2 1
Say you’re using a C scale,that would be G A G F E D C. Sing this against a C chord.
Or if you prefer solfeggio, | so la so fa |mi re do
Stop reading for a moment. Try to hum this phrase. If you can hear tunes really clearly in your head, you can skip the humming.
So far so good. That’s an eighth of a verse done already. This phrase ended on a C chord. Pretty often, that means the next one will end on a G chord. So start playing a G chord and put the stressed syllables on the first, third, and fifth notes of the G scale. I’m putting vertical lines to indicate where the musical bars begin and end.
But (D eve)ry (G song )has (D rhy) thm
| G A G F | E D C or | so la so fa | mi re do | D E G F | D G mi | re mi so fa | re so
Well, that fits the rules, but it's ugly. Let's add a new rule: a short stressed note doesn't have to be a chord note if it is followed by a long note that is a chord note. Let the last two notes be E and D, mi and re. That sounds better.
The contrast and interplay of musical phrases are what makes Western music interesting. In general, people like to be able to figure out what’s going to happen, and they like to be surprised. If every phrase is different, they will feel lost and blame it on you. Dumb melody. If every phrase is the same, they will get bored, even if you try to placate them with wonderful words and lots of performance values.
But before we make them think, we’ll help them feel at home and save ourselves some work. We’ll use the first phrase over. Bingo, or Shazam if you prefer, we have three lines for the price of two.
Now for the fourth line. Can we get away with using the second phrase again? Probably not. And it’s not time to wrap up the melody in a satisfying final resolution, because we’re only halfway through the verse. So it’s going to end on a G chord again, but let’s introduce some tension by starting with a different chord. As a general principle, chords like to resolve in one direction, the well-known circle of fifths. Here’s part of it: E -> A -> D -> G -> C -> F. So let’s try jumping to a D chord for half of the line, and finishing with a G chord. Picking notes that fit in the D chord, then the G chord:
| G A G F | E D C or | so la so fa | mi re do E | D E G F | E D mi | re mi so fa | mi re | G A G F | E D C | so la so fa | mi re do E | D E F# D | A G mi | re mi fi re | la so
At the end of line 4, that A on a stressed note isn’t part of the G chord, but we know we’re going towards a long G note and it just adds a little tension and echos what we did in line 2. (Truthfully, I wrote line 4 before changing line 2.)
Half done. Now we could get away with the first phrase again, but it’s time to add a bit more harmonic interest and we haven’t used the F chord yet. A whole line of F? F returning to C? Either will work. Then the sixth line will be C and G or F and G, we can re-use phrase 1 for the seventh line, and the wrap-up line will resolve to a C chord, so it will begin with a G chord. The melody comes from chord notes just as it did for the first phrases.
That’s a quick and dirty way to write a halfway decent tune. I'll recapitulate:
I'll put more about melody writing in an appendix, but there’s not time to do anything really fancy in a short workshop.
You’ve just written a song. Grab your guitar, collect an audience, and humiliate yourself. Oops. Writing it and learning it are two different processes. You WILL forget your own words. You WILL lose the melody. Another unwelcome learning experience.
If you have developed a process for learning songs, go through it with your new creation. At very least, sing it a few times until your fingers know where the changes are and your lungs know where to breathe. If you can possibly arrange to sing it in front of a few discerning but forgiving friends, do that next. If you have a fear and loathing of recording devices, try to record it, so you will have an experience of singing it while stressed. If some guitar lick that you have improvised while practicing seems to work well, write it down, or at least practice it separately.
What you want to accomplish, BEFORE you sing your song in public, is to bring it to a state where you can concentrate on the meaning and the audience. Every brain cell that you have to devote to remembering what comes next is another chance to divert attention from the song to the problems you are having with it.
You don't bother to finish your songs. You stop when you have words and a melody.
It’s unusual to write a song so well that it is perfect the first try. Since you wrote it, you are in a unique position to know where the unsatisfying parts are. You are also in the very worst place to notice what works and doesn’t work. Ideally you have lots of time after writing it to practice it, record it, go on to something else, and come back to it.
The single most useful thing you can do with a raw song is to expose it to critique from other songwriters. Join a songwriting group if you can find one that isn’t closed, or round up some other aspiring songwriters and form one. Ask somebody, anybody who doesn’t hate music and/or you. A second opinion will turn up things that the song doesn’t say that you thought it did, and things it seems to mean that you had no idea of. The beauty of a songwriting group is that you pay for this useful information by getting a chance to think about the strong and weak points of other people’s songs.
One really important thing to remember when polishing a song is that nothing is chiseled in stone. You occasionally will find that your very favorite lines don’t work with the rest of the song, and you can’t bear to throw them out. Fine, don’t throw them out, throw them in – to a new song where they will feel at home. You may have written too many verses, or tried to do too many things. Or maybe it will be a much better song if you add some new insight.
Some poetry writing guides will tell you that every word has to work, and your job is to throw out any word that isn’t pulling its weight. That doesn’t mean that every “and" and “the” has to go, or that you have to throw away good clear meaning to make room for juicy words.
Polishing takes time, and it’s probably not possible to do it all in one session, or before trying it out with an audience that isn’t all songwriting buddies. Practicing to polish your performance will sometimes help you notice the rough spots in your song.
The least of your worries, unless you want to make big bets against big odds. If you want to make a bit of money, make the best recording you can, and get good enough at performing that you can get gigs within driving distance. Sell recordings at gigs. If you want to make the world a better place, sing and lead songs at hospitals or on picket lines. That is pretty much the best you can do by yourself.
I'm leaving that in to remind myself that times change. Send your recording to CDBaby. There are dozens of dollars to be made in digital downloads, even if they never sell a single CD for you. Stick a couple of decent performance videos up on YouTube. I am writing this in 2011. No doubt, it will sound even more dated than the preceding paragraph in the 2020s.
The music industry concentrates its publicity on a small number of stars, and anybody who survives the pressures of stardom has to develop a cast-iron barricade against being bombarded with songs. There are people who are paid to listen to demos, and they do. About ten seconds per song, unless there is something that grabs them in the first ten seconds. Now, a little bit of thinking will show you that not many good songs are really grabbing in the first ten seconds.
If you sing around, you will accumulate music friends, and if you write well enough, eventually somebody who knows you will tell somebody else about one of your songs.
Still, it doesn’t hurt to register your songs with a performing rights organization like ASCAP or BMI or (in Canada) SOCAN. If you do get airplay, there’s a small but real chance that it will happen when that station is being sampled. And if you perform at a registered venue and provide a list of the songs you sing, you will probably collect a few dollars.
Free writing, or journaling, is the most useful process I’ve taken from any songwriting class. Simply sit down with lots of paper and the writing instrument of your choice and write as quickly as possible for a fixed amount of time. Don’t take time to correct anything after you write it or to think anything through before you write it. This sounds incredibly stupid, but it is the one-(wo)man equivalent of a brainstorming session. Even if you just write “This is boring” over and over for ten minutes, sooner or later your mind will realize that no amount of negative thinking is going to stop you, and something will sneak out. There have been cases where people with particularly rigid feelings have filled whole notebooks with pointless complaints, then eventually worn down their censors and begun to get in touch with what they really had to say.
When you sit down to write and can’t think of what to do, don’t bludgeon your brain with attempts to try harder without knowing what to try. Do ten minutes of free writing. Something will probably emerge. If not, try again the next time.
When you do have a topic in mind, journal about it. Go back after your time has expired, circle the most hopeful words and phrases, and journal about them. This is the best way there is to get your thoughts out on paper where you can look at them. Maybe you have to take an extra few minutes to transcribe your random scrawls into organized lists of things to use, but you’ll end up with more than if you started out being organized.
Here's a songwriting trick that Kristen Allen-Zito showed me. Get a very bad tape recorder and sing nonsense syllables to it. Then, a day or so later, listen to the bad recording and try to figure out what the words were. Cherry-pick phrases from your transcription and crystallize song ideas around them. Really bad tape recorders are hard to find and clunky to carry around, but you can make a good recording on your cell phone, load it into your computer, and run it through whatever sound editor you use until you can't understand it.
You may feel that you already criticize yourself plenty. Sorry: unless you are very unusual, what you do is attack yourself. No good critic would dream of heaping scorn and abuse over a work in the early stages. It would be, to use Fl!p Breskin’s favorite analogy, like scolding a baby for clumsiness while he or she is learning to walk. Your first job as a self-critic is to learn to turn off the pointless nastiness, and it isn’t easy. Free writing is a tool for getting around an untrained “critic”.
Once you have a bunch of stuff on paper, it is important to be able to notice how the process of organizing it is going. When you are making lines that scan and rhyme, it is extremely important to be able to notice when they don’t. None of this calls for feeling bad about yourself. Chant the lines while walking or tapping your finger. Use all the tools you need.
Your first song may not be much good, but you can’t write your hundredth song before the first ninety-nine. You’re walking a delicate line between learning to see the places that could be better, so you can make them better, and giving your inner hostile parrot a chance to eat your soul.
One good songwriting assignment is worth six months of psychotherapy. Writing songs can help you get in touch with what is bugging you, and that’s the first step in healing.
On the other hand, the first time I heard the phrase “singer-songwhiner”, I knew exactly what it meant. Seldom will a highly personal song written to help you heal yourself also be a song that a general audience will appreciate. There is a niche for feeling-sorry-for-myself songs, but there are better ways to entertain folks and change the world. Think about people who have had interesting lives and written interesting songs about it, but now they’re on the road singing those songs and all they can write about is what a drag it is to be on the road.
You want to know about real therapeutic songs, find out what works best when you sing in hospitals and rest homes.
There are several really good reasons to begin with the tune.
Consider the parody song. Start with somebody else’s song and write new words. You might preserve something of the organization and content of the original song, or do something completely different and just use the tune. Clive Gregson uses an interesting trick when he teaches songwriting. One student writes a new set of words to a Beatles tune (anything but Yesterday), then another takes these words and write a new tune for them. I’ve never seen anybody guess which Beatles tune the first writer started with.
Robert Burns wrote words to some of his favorite Scottish melodies so people would learn the melodies.
There are many wonderful songs in other languages. It is perfectly acceptable to sing them in the original language in front of an audience that doesn’t understand it, but if you can come up with a translation that fits the meaning and the tune, what a contribution!
Musical collaborators go either way, Gilbert before Sullivan and Hammerstein before Rodgers, but Rodgers before Hart. If you can round up a Rodgers or a Sullivan, skip the sections on writing melodies and get to work on your lyrics. If you *are* a composer of that level, you’re welcome to all of my lyrics.
A variation on starting with a tune is starting with a chord sequence. By far the most common set of shared chords is the twelve-bar blues, and there are hundreds of very different blues that all fit the same chords. Here are the chords for 12-bar blues in the key of E and in key-independent symbols.
E E E E I I I I F F E E IV IV I I B B(or A) E E V V(or IV) I I
Jazz people often write songs that fit the chords of existing standards. You can get the chords to a song you don't know from a jazz fakebook or from Rise Up Singing and try to write a melody that fits them.
In the main body of this guide, I touched very briefly on how to write melodies. Here’s a bit more.
A popular songwriting assignment is to assign the digits in your telephone number, or the pips on a handful of playing cards, to the notes of a scale. You use that as the starting point of a tune. This is technically “white noise” music, where there is no relationship between any note and any other note. It’s a good trick for coming up with phrases you’d never have written yourself, and it’s an interesting challenge to write the rest of a melody so that the white-noise phrase fits in and makes sense.
The other extreme of mechanically generated melody is starting somewhere and flipping a coin to see whether the tune goes up or down a note. That’s “brown noise”, from the Brownian motion of a smoke particle as it bounces off of air molecules. “Brown” melodies sound like really bad Gregorian chant.
It is certainly possible to generate more interesting melodies than either of the above, by writing a set of rules and flipping a coin to decide how to apply them. Computers are good at this. If you program a computer to find out what rules Bach or Beethoven or Bernstein seemed to follow, you can generate tunes that sound like Bach or Beethoven or Bernstein after a lobotomy.Here is a link to www.zekehoskin.com/tunegen.htm, a web page containg a simple tune generator which will make random ten-note tunes that are brown, white, or pink (halfway between).
Real music isn’t random. At every level – notes, bars, phrases, lines, stanzas, movements, symphonies – the composer is striving to provide patterns that change and repeat in ways that you can follow but still be surprised by. This generally means that every given level is a pattern of items on the smaller level, repeated and varied in a meaningful way. There are several ways to listen to the melody of a verse:
How do the phrases develop? I tend to think of a question phrase followed by a tentative answer, or maybe four question-answer phrase pairs with the second a bit less tentative and the last more definite. This is mostly but not altogether a matter of where the last note of each phrase fits into the scale of the tune and the chord that is being played at the time.
Listen to just the bass line, or just the chord sequence.
Very often, it's helpful to work with a motif, a short chunk of melody, maybe just three to five notes. Different phrases might contain the same motif, or variations of it.
Is there a harmony part? If you are singing by yourself, probably not. Writing for two singers is different. Each of the parts has to fit into the useful range of one of the singers. Maybe instead of a second part with the same rhythm as the main melody, you might write a counterpoint part. Listen a duet doing Suwanee River with Humoresque, or It’s A Sin To Tell A Lie with Cross My Heart And Hope to Die.
Are there places with no words? Some songs are completely full of words and it’s hard to hear where the singer has time to breathe. Others have short lines with long spaces between them. Maybe there is room for a tasteful instrumental fill in there, or maybe the instrument is noodling around and detracting from the song. Did you write out the tune for the fills, or leave room to ad-lib?
As an artist and a human being, you deserve the time and space to write music. You won’t get them if you don’t make it happen.
Find out what works for you. I can write lyrics on the bus, but melodies are spookier. At a pinch, I can work in a reasonably noisy house by wearing headphones playing the sound of a burbling brook. Walking in a quiet forest is much better, but harder to arrange. Your songwriting requirements are bound to be different from mine.
Like any skill, songwriting develops best with regular practice. An hour three times a week is way better than twelve hours once a month. See if you can arrange your life so you get that hour.
Some people find that carrying a small voice recorder wherever they go lets them grab odd moments of inspiration and opportunity. Others just break a lot of voice recorders and learn to record their ideas on paper.
Listen to a lot of music. When you hear a song you like, learn it. Learn all the words and all the notes and all the chords and all the dynamics. Listen to more songs by the same writer.
Read a lot of the kind of poetry that rhymes and scans.
Write rigidly patterned poems. If you can write a limerick and a sonnet, song verses will come easier.
Learn to play an instrument.