Suno
An AI music platform that turns text prompts into full songs with lyrics, vocals and instrumentation.
Suno reached 2 million paying subscribers and $300M ARR, with about 100 million people having made music on it. It started as a paid Discord bot, moved 90% of traffic to the web within 5 days, and kept the free plan at five creates a day. Growth came mainly from better models and features. Lesson: make the product better, and make shared content invite new people to create.
- 1Make a song you loveSharing it feels 10 or 100x better
- 2Share it anywhereText, WhatsApp, social posts, photo slideshows
- 3Listener hears, no loginPage makes clear it's made on Suno
- 4Listener riffs on itShared song solves the blank canvas
From Mikey Shulman, Suno cofounder, podcast interview
What was said about Suno
1Why better models, not growth tactics, drive Suno's growth
Every growth jump came from a new model or the products it enabled. Suno reached 2 million paying subscribers and $300M ARR, and Mikey Shulman says all inflections in users, revenue and subscribers came from making the product better, with model releases the biggest drivers.
The team tracks three things in a model: sound quality, song quality (catchy, memorable) and control (how users interact with it). Songs and control can always improve, so the lever does not run out. Quote: "basically nothing else ends up mattering."
Leave the paywall alone while the product changes every few months. The free plan has stayed at five creates a day since launch, about three years, while the company grew to $300M ARR.
Shulman argues against trading usage for revenue by moving the paywall, because capabilities change so fast that "any optimizations we might make to the free tier, um, are going to be stale in like two months." The aim is to raise both usage and revenue by improving the product.
New models go to paid users first, then reach the free tier. Each new model is released only to the paid tier, and the free model is upgraded later once the new one is cheaper to serve.
Shulman admits the heuristic isn't deeply thought out and may change.
2The shared song as the growth engine
Making a song yourself makes sharing it far more rewarding. Sending a great Spotify find makes you feel like a tastemaker; sharing a song you made amplifies that feeling by 10 or 100 times, per Shulman. Most new users arrive through a shared song.
People share by text, WhatsApp, social posts and slideshows of family photos. Suno does not dictate channels; it makes the path to a song people love easier, and "every time we make the journey to get to that piece of content easier, better, more enjoyable, we just see the sharing behaviors explode."
Keep the shared song page open but make creating obvious. Listeners hear a shared song without logging in, while the page makes clear it was made on Suno and that they can make one too.
Shulman believes roughly 90% of people who receive a shared song should be convertible into making one, but this is his intuition, not a measured result. He calls the balance an art where each improvement pays off. Quote: "The biggest one is making it obvious."
A shared song solves the blank canvas problem for new users. Asking a newcomer to create from nothing stalls them; a shared song gives built-in inspiration to riff on, adapt, cover or sample.
Nearly 50% of first-time users hit the free limit, and about 100 million people have made music on Suno. The team plans to build more around starting from an existing song.
3Getting users to feel a song is their own
- 1Type a short promptSix words alone feels gimmicky
- 2Edit the lyricsClick into an okay result
- 3Write your own lyricsHooked on creating and iterating
- 4Sing or use VoicesYour voice from about 15 seconds
- 5Long sessions in StudioOwned, good songs drive retention
From Mikey Shulman, Suno cofounder, podcast interview
Retention hinges on making a song that feels yours in early sessions. Shulman's model: users retain if they made a song that felt like their own and that was good, and the first one or two sessions matter most.
Ownership is a product experience problem (more than typing six words into a box); quality is largely a model problem. Users who get the magical moment retain and likely subscribe.
Editing steps turn a passive prompt into creative ownership. Users who get an okay result click in to edit lyrics, then start writing their own, which hooks them on creating and iterating.
Singing into the microphone builds ownership too, and the path leads toward long sessions in the Studio pro tool. A song from six typed words alone feels gimmicky.
Personal features like Voices and My Taste make users attached. Suno 5.5 shipped Voices (your own voice from about 15 seconds of singing), My Taste (an editable memory of your preferences) and Custom Models (upload your own music as context). Voices was among the most requested features, possibly the most.
The goal is a personal producer who gets you, for everyone. Quote: "When things feel personal, you fall in love with them."
Position the product around creating, not listening. Suno deliberately is not an AI Spotify, since Spotify already has more music than anyone can hear; it sells creative entertainment.
Shulman frames value as a pyramid of utility, entertainment, then fulfillment at the top: "You can get entertainment anywhere. The top of the pyramid is fulfillment." He also stresses being on the phone, because "creativity strikes at weird times" and the phone is usually with you.
4Launching as a paid Discord bot, then moving to the web
A crude Discord bot got Suno in front of users with no UI. The first product, modeled on Midjourney, made 12.5 seconds of poor audio from a one-sentence prompt, yet people played with it for hours.
Discord removed the need to build a UI or login, and users used the product, gave feedback and told others in the same place. Early feedback said the music was bad, but making it was enjoyable, matching the founders' own experience.
Charging from day one showed who really valued the product. Suno charged on the Discord bot from the start, against the usual advice to grow first, and many people signed up in the first month.
Early on the team could talk to every person who hit the paywall and didn't pay and ask why. Shulman notes he has no counterfactual on whether it was the right call.
A basic web app pulled 90% of traffic off Discord in 5 days. The December 2023 web app lacked the bot's full functionality, yet about 90% of traffic moved over within 5 days and infrastructure crashed a few times.
Shulman expected to stay on Discord long, as Midjourney did, and says he was wrong; his cofounders pushed for the web. Nobody should need to recover a Discord password to use a music machine. The launch fell around Thanksgiving, when people were home with family.
5Bark, timing and choosing music over speech
Suno started when open-source text work let it focus on audio. Shulman says six months earlier there wasn't enough open-source text work to reuse, and six months later competition started to creep in.
Audio was a neglected "second-class citizen" in AI, so borrowing from the text community let a small team spend its energy on audio-specific problems and ship before others.
Open sourcing Bark validated demand for expressive audio before monetizing. Bark, an inefficient text-to-audio model too costly to commercialize, quickly gathered GitHub stars and a developer community.
It made word errors but sounded emotional and lifelike, and could even whisper, unlike speech systems tuned for word error rate and speed. That showed subjective quality mattered more than the standard metrics.
Pick what's only possible now over doing the old application faster. Many urged Suno to build a speech company as the bigger, simpler market, but it chose music and artistic uses.
Speech players like ElevenLabs and Speak compete on accuracy, voice variety and coaching. Quote: "how do I do the same application faster? I think often misses some of the beauty that comes with like doing something that was only literally possible literally today."
6Suno's TikTok creator program
Suno lets AI-music hate in TikTok comments drive engagement. Suno's TikTok comments are full of people saying they hate AI music, which lifts engagement and watch time, and the app hit #1 in the App Store music category.
The advice for polarizing products is to provoke a reaction while still showing value, without putting other brands down. Quote: "You want to get under people's skin a little bit, but also show value."
Run about one creator manager per 10 creators. Suno has around eight creator managers who coach creators rather than make content themselves.
The recommended ratio is one manager to 10 creators, 15 at most; under five wastes money and over 15 gets lost. Managers also recruit new creators and cut weak ones fast, since low-ticket videos mean high churn.
7Suno as a tool for ad testing
Marketers use Suno to test many music tracks on one ad. Phil Carter suggests layering several Suno tracks onto one 30-second ad featuring real people to see which performs best, as part of testing hundreds of creatives a month.
He warns against fully AI-generated ads because of quality, FTC risk on fake testimonials and consumer backlash.
How Suno Grew to $300M in ARR by Democratizing Music Creation
Grew to 2 million paying subscribers and $300M ARR, with about 100 million people having made music on it. It started as a paid Discord bot in 2023, moved 90% of traffic to a web app within 5 days, and keeps the free plan at five creates a day. Growth inflections come from better models and products, such as Voices, My Taste and custom models in v5.5.
Suno lets AI-music hate in TikTok comments drive engagement. Suno's TikTok comments are full of people saying they hate AI music, which lifts engagement and watch time, and the app hit #1 in the App Store music category.
The advice for polarizing products is to provoke a reaction while still showing value, without putting other brands down. Quote: "You want to get under people's skin a little bit, but also show value."
Run about one creator manager per 10 creators. Suno has around eight creator managers who coach creators rather than make content themselves.
The recommended ratio is one manager to 10 creators, 15 at most; under five wastes money and over 15 gets lost. Managers also recruit new creators and cut weak ones fast, since low-ticket videos mean high churn.
Founder of CAA Michael Ovitz: "I Just Knew I Was Going to Outwork Every One of Them"
Ovitz says Suno and Udio take other people's IP, even 10 seconds of it, to build custom songs. Its effect on music is unresolved, though piracy offers a precedent.
Every growth jump came from a new model or the products it enabled. Suno reached 2 million paying subscribers and $300M ARR, and Mikey Shulman says all inflections in users, revenue and subscribers came from making the product better, with model releases the biggest drivers.
The team tracks three things in a model: sound quality, song quality (catchy, memorable) and control (how users interact with it). Songs and control can always improve, so the lever does not run out. Quote: "basically nothing else ends up mattering."
Leave the paywall alone while the product changes every few months. The free plan has stayed at five creates a day since launch, about three years, while the company grew to $300M ARR.
Shulman argues against trading usage for revenue by moving the paywall, because capabilities change so fast that "any optimizations we might make to the free tier, um, are going to be stale in like two months." The aim is to raise both usage and revenue by improving the product.
New models go to paid users first, then reach the free tier. Each new model is released only to the paid tier, and the free model is upgraded later once the new one is cheaper to serve.
Shulman admits the heuristic isn't deeply thought out and may change.
Making a song yourself makes sharing it far more rewarding. Sending a great Spotify find makes you feel like a tastemaker; sharing a song you made amplifies that feeling by 10 or 100 times, per Shulman. Most new users arrive through a shared song.
People share by text, WhatsApp, social posts and slideshows of family photos. Suno does not dictate channels; it makes the path to a song people love easier, and "every time we make the journey to get to that piece of content easier, better, more enjoyable, we just see the sharing behaviors explode."
Keep the shared song page open but make creating obvious. Listeners hear a shared song without logging in, while the page makes clear it was made on Suno and that they can make one too.
Shulman believes roughly 90% of people who receive a shared song should be convertible into making one, but this is his intuition, not a measured result. He calls the balance an art where each improvement pays off. Quote: "The biggest one is making it obvious."
A shared song solves the blank canvas problem for new users. Asking a newcomer to create from nothing stalls them; a shared song gives built-in inspiration to riff on, adapt, cover or sample.
Nearly 50% of first-time users hit the free limit, and about 100 million people have made music on Suno. The team plans to build more around starting from an existing song.
Retention hinges on making a song that feels yours in early sessions. Shulman's model: users retain if they made a song that felt like their own and that was good, and the first one or two sessions matter most.
Ownership is a product experience problem (more than typing six words into a box); quality is largely a model problem. Users who get the magical moment retain and likely subscribe.
Editing steps turn a passive prompt into creative ownership. Users who get an okay result click in to edit lyrics, then start writing their own, which hooks them on creating and iterating.
Singing into the microphone builds ownership too, and the path leads toward long sessions in the Studio pro tool. A song from six typed words alone feels gimmicky.
Personal features like Voices and My Taste make users attached. Suno 5.5 shipped Voices (your own voice from about 15 seconds of singing), My Taste (an editable memory of your preferences) and Custom Models (upload your own music as context). Voices was among the most requested features, possibly the most.
The goal is a personal producer who gets you, for everyone. Quote: "When things feel personal, you fall in love with them."
Position the product around creating, not listening. Suno deliberately is not an AI Spotify, since Spotify already has more music than anyone can hear; it sells creative entertainment.
Shulman frames value as a pyramid of utility, entertainment, then fulfillment at the top: "You can get entertainment anywhere. The top of the pyramid is fulfillment." He also stresses being on the phone, because "creativity strikes at weird times" and the phone is usually with you.
A crude Discord bot got Suno in front of users with no UI. The first product, modeled on Midjourney, made 12.5 seconds of poor audio from a one-sentence prompt, yet people played with it for hours.
Discord removed the need to build a UI or login, and users used the product, gave feedback and told others in the same place. Early feedback said the music was bad, but making it was enjoyable, matching the founders' own experience.
Charging from day one showed who really valued the product. Suno charged on the Discord bot from the start, against the usual advice to grow first, and many people signed up in the first month.
Early on the team could talk to every person who hit the paywall and didn't pay and ask why. Shulman notes he has no counterfactual on whether it was the right call.
A basic web app pulled 90% of traffic off Discord in 5 days. The December 2023 web app lacked the bot's full functionality, yet about 90% of traffic moved over within 5 days and infrastructure crashed a few times.
Shulman expected to stay on Discord long, as Midjourney did, and says he was wrong; his cofounders pushed for the web. Nobody should need to recover a Discord password to use a music machine. The launch fell around Thanksgiving, when people were home with family.
Suno started when open-source text work let it focus on audio. Shulman says six months earlier there wasn't enough open-source text work to reuse, and six months later competition started to creep in.
Audio was a neglected "second-class citizen" in AI, so borrowing from the text community let a small team spend its energy on audio-specific problems and ship before others.
Open sourcing Bark validated demand for expressive audio before monetizing. Bark, an inefficient text-to-audio model too costly to commercialize, quickly gathered GitHub stars and a developer community.
It made word errors but sounded emotional and lifelike, and could even whisper, unlike speech systems tuned for word error rate and speed. That showed subjective quality mattered more than the standard metrics.
Pick what's only possible now over doing the old application faster. Many urged Suno to build a speech company as the bigger, simpler market, but it chose music and artistic uses.
Speech players like ElevenLabs and Speak compete on accuracy, voice variety and coaching. Quote: "how do I do the same application faster? I think often misses some of the beauty that comes with like doing something that was only literally possible literally today."
The AIGC Ad Creatives Crushing It vs. The Ones Flopping
Likely used to generate the music and lyrics in the dancing raccoon ad, which the host took for a real song. Reface tests many tracks because sound drives performance more than expected.
Marketers use Suno to test many music tracks on one ad. Phil Carter suggests layering several Suno tracks onto one 30-second ad featuring real people to see which performs best, as part of testing hundreds of creatives a month.
He warns against fully AI-generated ads because of quality, FTC risk on fake testimonials and consumer backlash.
I Studied 1,000,000 UGC Ads, This Is What Will Grow Your App
Suno leans into hate for AI music in its TikTok comments to lift engagement and watch time, and hit #1 in the App Store music category. It runs about eight creator managers who also coach and do not create content.
The AI Growth Playbook for Subscription Apps – Phil Carter, Elemental Growth
Can layer several music tracks onto one 30-second ad to see which performs best.
How Wispr Flow Uses AI to Save Professionals Hours Every Day
Tanay's favorite little-known AI app, which he calls maybe the most underrated AI product ever. He listens to songs he makes there more than Spotify.
How Runna’s Laser Focus Led to Its Successful Acquisition by Strava
Runna uses it to create music for ads, avoiding terrible royalty-free tracks and making reruns of winning ads easier.
Unconventional product lessons from Binance, N26, Google, more | Mayur Kamat (CPO at N26)
Mayur uses it to turn his Hindi poems into songs and finds the core tech magical. He would change its onboarding and growth product.
More in Music & Audio
Music & AudioSpotify
A music, podcast and audiobook streaming app with a freemium model and personalized recommendations.
26 sources · 61 tips
Music & AudioAnchor
A podcast creation and hosting app, later renamed Spotify for Podcasters.
1 source · 10 tips
Music & AudioBeats
A headphone and audio brand co-founded by Jimmy Iovine and Dr. Dre.
2 sources · 6 tipsBeats Music
A streaming service built from the MOG purchase that became Apple Music.
1 source · 2 tipsPhonograph
Edison's device for recording and replaying the human voice.
1 source · 2 tipsFirst Roots Media
A startup letting independent musicians sell or share music and track how far fans spread it.
1 source · 1 tipsimeem
A music discovery and social media startup founded by Dalton Caldwell.
1 source · 1 tipsRain Sounds
A sleep and ambient sounds app made by the guest.
1 source · 1 tips