How to Use AI to Create Music (That Actually Sounds Good)

You can create a full song with AI in under a minute: describe the genre, mood, instruments, and tempo in one prompt, generate two versions, keep the best, and edit the rest. Suno’s v6 models — launched September 9, 2026 — are the current default for beginners, Udio is the experimental alternative with a serious catch, and ElevenLabs Music v2 is the best option if you already pay for ElevenLabs. This guide walks through the whole workflow, the real prices, and the licensing rules most tutorials skip.

How I tested this: I verified every step, price, and limit in this guide against the official Suno, Udio, and ElevenLabs pages on November 11, 2026, and linked each claim to its source. I did not run a paid generation session for this guide, so the prompt workflow below follows the official documentation and the documented behavior of the current v6 models — not a live screen recording.

Why learn AI music now?

AI music stopped being a novelty in September 2026. On September 9, Suno launched v6 — three new models (v6, v6-wild, and v6-mini) trained on licensed music from Warner Music Group, BMG, and Believe, replacing every older model outright. The free tier got a real model of its own: v6-mini runs on every account, including free ones.

Demand followed. Suno drew about 550,000 US searches in September 2026, up 22% year over year, and its US web visitors grew from 1.87 million to 3.21 million in a year — up 72% and within 2% of its 2026 peak, according to New Engen’s analysis of Similarweb data. Suno also controlled 69.2% of measured AI music-generator traffic in the twelve months ending September 2026, per OneLittleWeb’s traffic dashboard. In other words, this is no longer a side experiment: it is the default way beginners make music with AI, and the tools are finally good enough that the results sound like finished songs rather than demos.

Which AI music tool should you use?

Three tools cover nearly every beginner use case. Pick by what you need the song for, not by which one has the longest feature list — the licensing terms matter more than the feature lists, as the table shows.

My pick for most people: Suno. It has the largest community, the clearest free tier, and the only paid plans that include both downloads and commercial use rights. Udio sounds great, but since its Universal Music Group partnership on October 29, 2025, downloads of audio, video, and stems are disabled for all users, and its terms bar commercial use on every plan — so treat Udio as a sketchpad, not a distribution tool. I compared the two in depth in my Suno vs Udio head-to-head if you want the full side-by-side.

How to create your first song with Suno

Suno’s web app at suno.com is where nearly every beginner should start. The whole first song takes about five minutes, and the free plan needs no credit card.

Step 1: Create a free account

Go to suno.com and sign up with Google, Discord, or email. The free plan gives you 50 credits a day and access to v6-mini, the free model. You cannot download songs or use them commercially on the free plan — you can listen and share them by link — so treat your first tracks as practice.

Step 2: Start in Simple Mode

Click Create and describe your song in one plain sentence, the way you would describe it to a friend. Suno writes the lyrics and picks the style for you. Example: “An upbeat indie-pop song about a road trip at sunrise, female vocals, catchy chorus.” Click Create and Suno generates two versions in about a minute. Listen to both all the way through — AI hides its mistakes in the middle of a track, not the first ten seconds.

Step 3: Switch to Custom Mode for control

Simple Mode is a slot machine. Custom Mode is where the real workflow lives. Toggle it on and you get three fields that matter:

Style of Music: genre, instruments, mood, tempo, and vocal character — never the song itself.

Lyrics: your own words, or leave it blank and let Suno write them.

Title: what the track is called in your library.

Use structure tags on their own lines inside the lyrics box to control the arrangement: [Verse], [Chorus], [Bridge], and [Outro]. Suno’s v6 models follow these tags reliably, which is the single biggest difference between a song that sounds structured and one that rambles.

Step 4: Generate two versions and pick one

Every generation produces two variations. Do not ship the first one. Listen to both fully, keep the stronger take, and delete the other — your library fills up fast on a free plan. If neither works, change exactly one thing in the prompt (the tempo, one instrument, or one lyric line) and generate again. Changing five things at once teaches you nothing about what fixed it.

Step 5: Extend, edit, or download

Happy with a section but want a longer track? Use Extend to continue from a timestamp, or Replace Section to regenerate one part without touching the rest. On Pro, you can separate a track into stems (vocals, drums, bass, melody) for mixing elsewhere; Premier adds Suno Studio, a browser DAW with MIDI, effects, and automation. Downloads are capped — 20 songs a month on Pro, 60 on Premier, none on Free — so download only the takes you actually intend to use.

How to write a prompt that sounds like music, not noise

A good music prompt has six slots. Fill all six and the model has almost nothing left to guess:

Genre + era: “lo-fi hip-hop,” “80s synthwave,” “acoustic folk”

Instruments: “mellow electric piano, soft drums, warm bass”

Mood: “relaxed,” “euphoric,” “melancholic”

Tempo: “72 BPM” beats “slow”

Vocal character: “soft female vocals” or “no vocals”

Structure or length: “30-second intro that fades out cleanly”

Here is a worked example I would use for a podcast intro — the most common beginner project, and one where a short instrumental is exactly right:

Once the intro is ready, it slots into the full workflow in our guide on how to use AI to create a podcast, which covers scripting, AI voices, editing, and publishing the episode.

Style of Music: “Lo-fi hip-hop podcast intro, 30 seconds, mellow electric piano over soft drums, warm vinyl crackle, relaxed 72 BPM, fades out cleanly, no vocals”
Lyrics: leave blank — an instrumental needs no lyrics box at all.
Title: “Morning Static — Podcast Intro”

And a full-song example, if you want vocals:

Style of Music: “Upbeat indie-pop, jangly electric guitar, driving drums, bright female vocals, 122 BPM, anthemic chorus”
Lyrics:
[Verse]
Packed a bag before the sunrise
Two coffees and a paper map
[Chorus]
We are going where the highway bends
Chasing light with my oldest friends

Two rules that save the most time: describe the sound, not a famous artist (“in the style of Taylor Swift” gets you a refusal or a bad imitation — “bright pop, confessional lyrics, 120 BPM” gets you a usable song), and keep lyrics short. Long verses with no structure tags are where AI vocals start mumbling.

What if you want lyrics?

You have three honest options, in order of quality. Best: write your own chorus — eight lines in your own words — and let the AI fill in the verses around it, because your words are the part no model can fake. Good: let Suno write the lyrics, then edit the specific lines that sound generic, and fix any names or places it invented. Weakest: paste in lyrics you did not write. Never paste copyrighted lyrics into a generator — the model may reproduce them, and the result is not yours to publish.

If your project needs spoken words rather than sung ones — a podcast intro with a voiceover, for example — pair the track with an AI voice tool instead of forcing the music model to speak. My roundup of the best AI voice generators covers the options that work well alongside a Suno track.

What does AI music cost?

Prices below are from the official Suno, Udio, and ElevenLabs Music pricing pages, checked November 11, 2026. Suno’s prices are shown with annual billing first; monthly billing costs more.

The practical takeaway: a Suno Pro subscription at $8 a month on annual billing is the cheapest path to music you can actually download and use in a monetized video. ElevenLabs Music is the better deal only if you already pay for ElevenLabs for voices — Music v2 is included in every ElevenLabs plan, and a 3-minute song costs 900 credits. Udio’s paid plans buy you more generations, not more rights: paying does not unlock downloads or commercial use.

Can you publish or sell AI-generated music?

Yes — but only on the right plan, and the rules changed in 2026, so old tutorials are dangerous here. Read the terms on the day you download, not the day you generate.

Suno: paid Pro and Premier plans include commercial use rights, and those rights survive cancellation for songs you made while subscribed. The free plan has no commercial rights at all. Suno’s v6 models were trained on licensed music from Warner Music Group, BMG, and Believe — but Universal Music Group and Sony Music are not part of that deal, and they filed a second lawsuit against Suno on September 18, 2026, so the legal picture is still moving.

Udio: do not plan a release around Udio. Its terms, last revised November 12, 2025, say Udio and its licensors own the output, limit you to personal, non-commercial use, and prohibit downloading songs or posting them to Spotify, YouTube, TikTok, or SoundCloud. Downloads have been disabled on every plan since October 29, 2025.

ElevenLabs Music: paid plans are licensed for commercial use, and ElevenLabs announced an AI music agreement with Universal Music Group in September 2026. The free plan is personal use only.

Copyright: the US Copyright Office’s January 2025 report concluded that purely AI-generated material, with no human authorship, cannot be copyrighted. Your lyrics, your arrangement choices, and your edits are the human contribution that can be protected — a raw one-prompt generation is not.

One more caution that applies to every AI tool, not just music: if a track will appear in a video, a podcast, or an ad, keep a local copy of the terms in force on the day you downloaded it. Terms change overnight — Udio’s did, in October 2025 — and the version that matters is the one that applied when you got the file. If you want to check whether a piece of audio or video floating around online is AI-generated, my guide to SynthID Detector shows how Google’s watermark check works.

What to do with your first song

A song sitting in Suno’s library helps nobody. The three uses that actually teach you something are: a podcast or video intro (30 seconds, instrumental, no vocals to go wrong), background music for a short video (pair it with one of the best AI video generators and our guide to how to use AI to create a video, and you have a finished clip in an afternoon), and a demo for a real musician — AI is genuinely useful as a sketchpad you hand to a human to finish. What it is not, yet, is a replacement for a real performance: listeners forgive a lot in a demo and very little in a released track.

5 mistakes beginners make with AI music

One vague prompt, one generation, publish. Generate at least two versions of everything and listen to both fully.

Prompting a famous artist by name. Describe the sound instead — genre, tempo, instruments — and you will get a usable song instead of a refusal or a bad imitation.

Ignoring the plan’s limits. Free Suno cannot download or go commercial; Udio cannot do either on any plan. Check before you build a project around a track.

Long lyrics with no structure tags. Without [Verse] and [Chorus] markers, AI vocals ramble and mumble.

Skipping the full listen. Mispronounced words, tempo drift, and a chorus that never arrives all hide in the middle of the track.

FAQs