Quick answer
Screenshot the conversation, use your phone's built-in Live Text (iPhone) or Google Lens (Android) to copy the words out of the image, then paste them as lyrics into an AI music generator like Suno and pick a genre. The comedy comes from mismatch: mundane texts sung with complete sincerity in a genre that treats them as tragedy. Keep the thread short, keep the typos, and make sure the last line is the punchline.
You take a real conversation, usually a boring one, and you have an AI sing it. Not adapt it, not write a song about it. Sing the actual messages, word for word, typos included, with total sincerity.
- Jess
is everyone alive
- Maya
define alive
tara do you remember facetiming your ex
i did NOT facetime marcus
- Priya
tara do you remember him picking up
WHAT
- Maya
tara do you remember crying for forty minutes
oh my god
- Priya
tara do you remember asking to speak to his mum
He Picked Up
90s Power Ballad
Tara do you remembercrying for forty minutes…asking to speak to his mum
What This Trend Actually Is
The result works because of a gap. The words are somebody asking for a ride home or complaining about a shift they picked up. The music is treating those words as though they were the last thing a dying soldier wrote. Nobody in the recording seems to notice the mismatch, and that is the joke.
It started on April 11, 2026, when TikTok user @anotherdayatthestoddards turned her daughter's texts asking for Starbucks into an emo song. It passed 8.6 million views in under three weeks, and the format was fully formed on the first try. Every version since has been a variation on that one idea.
What You Need
Two things, both of which you probably already have.
- A conversation worth using. Not a dramatic one. A boring one with a small amount of friction in it.
- An AI music generator that takes custom lyrics. Suno is what most people use, and it shipped a screenshot-to-song feature during the trend. Its US downloads quadrupled week-on-week and it briefly hit number one on the music charts of both the US and UK App Stores, which tells you roughly how many people were making these at once.
You do not need an editor, a microphone, or any music knowledge.
The Steps
Step 1: Screenshot the conversation
Get the messages you want in one or two screenshots. Fewer is better. If you are already scrolling to capture it all, the thread is too long (more on that below).
Step 2: Pull the words out of the screenshot
You do not type these by hand. Your phone reads them for you.
- iPhone: open the screenshot in Photos, long-press directly on the message text. Live Text kicks in and you get the normal selection handles. Select all, copy.
- Android: open the screenshot and tap the Google Lens button, then select the text and copy.
- Suno: upload the screenshot directly and it does the reading step for you.
Do not clean up what you copy. Lowercase, missing apostrophes, autocorrect wreckage, the stray "k" as a full reply: all of it stays. Those artifacts are the proof that this was a real conversation and not a bit you wrote afterwards. A singer performing "ok well im nlt doing that" with a straight face is the entire product.
Step 3: Trim to the good part
This is the step almost nobody does and it is the one that decides whether the clip works.
Cut down to six to twelve messages. You want one clear line of escalation: a request, a refusal, a push, a resolution. If your thread does not have that shape, find one that does. Delete side tangents, delete the pleasantries at the top, delete anything that does not move the small conflict forward.
End on the funniest line. Not the last thing that was said, the funniest thing that was said. You are allowed to stop the lyrics early.
Step 4: Paste as lyrics and pick a genre
Paste the trimmed messages into the lyrics field. Keep the alternating speakers on separate lines. Then choose your genre, which is the second half of the joke and gets its own section below.
You can optionally bracket sections the way Suno expects, like [Verse] and [Chorus], and put the single most absurd line in the chorus so it repeats. This is worth doing. The repetition of one mundane sentence is reliably funnier than hearing it once.
Step 5: Generate, then generate again
Take a second and third pass. Vocal delivery varies a lot between generations on the same lyrics, and the difference between a version that lands and one that does not is usually just how committed the singer sounds. Pick the take where the voice sounds like it genuinely believes what it is singing.
Step 6: Post it with the screenshot on screen
The visual matters more than people expect. Viewers need to see the real messages while hearing them sung, otherwise there is no gap for the joke to live in. Put the screenshot on screen, scrolling in time with the vocals if you can be bothered, and let the audience read along.
Picking the Genre
The genre is not decoration. It is the punchline delivery mechanism. You are picking the style that will take your messages far more seriously than they deserve.
a song about accidentally liking a photo from 2023 while going through someone's instagram at 2am
Three Years Deep
Emo Ballad
I was only looking…and my thumb betrayed me,three years deep.
| Genre | What it does to the words | Best for |
|---|---|---|
| Emo / pop-punk | Frames minor inconvenience as personal betrayal | Teenagers asking for things, refused favours |
| Gospel | Treats the request as a matter of faith and deliverance | Anything involving asking a parent for something |
| Screamo | Total emotional collapse over nothing | Short, blunt, hostile messages |
| Opera | Absurd grandeur, every syllable held too long | Very short threads, one-word replies |
| 90s power ballad | Sincere heartbreak, key change and all | Mundane domestic negotiation |
| Country | Turns pettiness into hard-won folk wisdom | Complaints, grievances, work drama |
The rule underneath the table: the more emotional weight the genre carries by default, the less your messages need to have. A group chat about the morning after a night out is not inherently funny. Sung as a 90s power ballad called "He Picked Up," with a key change on the chorus, it is.
If you cannot decide, run the same lyrics through two genres and compare. It costs one extra generation and the difference is usually obvious within ten seconds. Emo and gospel are the safest opening bets because both are instantly recognisable within a couple of bars, which matters when you have three seconds to hold someone.
Why Some of These Flop
Most attempts at this trend die quietly. The reasons are consistent enough to list.
The thread is too long. This is the number one killer. A forty-message argument produces a song with no shape, no chorus, and no punchline, just a wall of sung text that people scroll past at eleven seconds. Short threads beat long ones every single time. If you have to choose between including a funny line and keeping it short, keep it short.
There is no escalation. A flat exchange where nothing builds gives the song nowhere to go. You want a small arc: someone wants something, someone else resists, it gets slightly worse, it ends. One arc, not three.
The last line is not the punchline. Songs are judged on how they end. If your final lyric is a logistical detail, the whole thing deflates. Reorder or trim so the funniest, flattest, most anticlimactic line is the one the singer holds the note on.
The typos got cleaned up. Corrected grammar makes the lyrics read as written comedy, and written comedy about texting is a much older and much less interesting genre than this one.
The genre matches the content. Sad texts sung as a sad song is not a joke, it is just a sad song. The mismatch is the entire mechanism. If the messages are already emotional, either pick a genre that is wildly inappropriate to them or pick a different conversation.
The screenshot is not visible. People need to read the real thing and hear the fake thing at the same time. Audio alone is half the joke.
It is 60 seconds long. These work at 15 to 30 seconds. The concept is understood in the first four bars, and everything after that is decay.
A Note on Using Other People's Messages
Ask first. Not a legal caveat, a practical one.
The versions of this trend that go well almost always involve someone who is in on it, usually a family member. The versions that go badly involve someone finding out that their private messages are being sung to strangers. Blur surnames, use first names or nicknames, and crop out anything identifying. If you would not read the thread aloud to the person's colleagues, do not turn it into a song.
Where Starrd Fits (Honestly)
Starrd is an AI video app: you upload a selfie and it puts you inside a cinematic scene. A song feature is in development and not launched yet, so there is nothing here for you to go and use today on the audio side, and I am not going to pretend otherwise.
The reason we are building it is the gap in the trend itself. Suno gives you a great piece of audio, and then you are on your own for the part TikTok actually needs, which is something to watch. Our angle is the video half. That is not available yet.
What does exist right now is the video template library, which is what Starrd already does well. If that is useful to you, it is at www.getstarrd.app. If it is not, the method above works completely without us, which is the point.
Sources and Further Reading
- Texts to Song Trend on Know Your Meme: the origin entry, including the April 11 video that started it.
- Rolling Stone on the TikTok AI text-to-song trend: how the format spread and what creators make of it.
- Music Ally: Suno is riding the wave of TikTok's new text-to-song trend: the download numbers and the App Store chart run.
Frequently Asked Questions
How do I turn my texts into a song? Screenshot the conversation, copy the words out with Live Text on iPhone or Google Lens on Android, paste them as lyrics into an AI music generator, and pick a genre that clashes with the content.
Where did the trend come from? TikTok user @anotherdayatthestoddards, on April 11, 2026, turning her daughter's texts asking for Starbucks into an emo song. It passed 8.6 million views in under three weeks.
What app do people use? Mostly Suno, which added a screenshot-to-song feature during the trend and briefly became the number one music app on the US and UK App Stores.
Do I have to type the messages out? No. Your phone's built-in text recognition reads them straight out of the screenshot, and Suno accepts the screenshot directly.
Which genre should I pick? Whichever one takes your messages far more seriously than they deserve. Gospel, screamo, opera, and 90s power ballad are the reliable four.
Why do most of these flop? The thread is too long. Six to twelve messages, one escalation, funniest line last.
Should I fix the typos? No. They are the evidence that the conversation was real.
Do I need to label it as AI? Yes, on every major platform.
Related Reading
- Viral AI Video Trends of 2026: what else is climbing right now.
