AI can make a whole song for you, with a voice singing the words. Type "make me a folk song about my dog's first trip to the beach" into a music generator and it writes lyrics, composes a melody, and returns a track with a sung vocal and instruments. You can also paste lyrics you wrote and have them sung. The tools differ on length, on whether you can download the file, and on what you may do with it afterward. Below: what each one does, the steps, the free routes, and who owns the result.
What you get when you ask AI to "make me a song"
An AI song generator is a tool that turns a written description, and optionally your own lyrics, into a finished audio track. You don't record, play or sing anything. As of September 2026, the consumer tools can do the following:
- Write the words for you. Describe the subject and leave the lyrics out. Google's help page says Gemini creates both the music and the lyrics from your description.
- Sing lyrics you wrote. Gemini accepts your own text marked with structure tags such as [Verse 1]. Toybox AI's Music Generator has a Lyrics box that holds up to 5,000 characters.
- Make instrumentals. Gemini lets you choose vocals or an instrumental, which suits background music under a slideshow or a video.
- Follow a style you describe. Google's prompt tips for Gemini list genre and era, tempo and rhythm, instruments, and the voice's gender, texture and range.
- Start from a photo. In the Gemini app you can upload a photo, and Gemini composes a track to fit the scene.
- Sing in other languages. ElevenLabs' music page names Spanish, French, German and Japanese alongside English.
You steer the melody only with words, so name the rhythm, instruments and voice. "A slow waltz with accordion and a warm baritone voice" tells a generator far more than "a happy song".
Can ChatGPT make songs?
OpenAI's list of ChatGPT features doesn't include a way to make music or songs you can play. The ChatGPT capabilities overview covers writing, search, deep research, image generation, file uploads, data analysis and voice conversations, and music or song generation isn't among them. ChatGPT handles the parts of a song that are words, and the audio has to come from a music generator.
That still makes it useful. Ask it for lyrics with section labels and a line length, then paste them into a tool that sings your lyrics. A request to start from: "Write lyrics for a country song about my brother Sam finishing his first marathon after two years of training. Label the sections [Verse 1], [Chorus], [Verse 2], [Chorus], [Bridge], [Chorus]. Keep every line under 10 syllables and make the chorus easy to sing along to." Read the draft out loud and shorten any line you can't say in one breath.
Can Gemini make music?
Yes. The Gemini app makes songs with Google's Lyria 3.5 music model, with vocals or as an instrumental, in tracks of up to 3 minutes. Google's help page lists three requirements: you're 18 or older, you're signed in with your own Google Account or a work or school account Google supports, and Keep Activity is turned on. These are Google's steps:
- Open the Gemini app or the Gemini web app, and switch the model to Pro. Google says a full track needs it.
- Tap "Upload & tools" at the bottom of the box where you type, then "Create music".
- Optionally, upload images or videos for context.
- Use the drop-downs to pick the length (short is about 1 minute, full is about 2 to 3 minutes), vocals or none, and a genre.
- Type your prompt. Google suggests starting with a verb such as write, compose or create.
- Tap Download at the top right of the track's image, then choose the video with cover art (MP4) or the audio file alone (MP3).
Google lists music generation for Gemini users with or without a paid Google AI plan. Your usage allowance resets on a 5-hour cycle, capped by a weekly limit. As of September 2026, Google AI Plus ($4.99 a month) and Google AI Pro ($19.99 a month) raise Gemini's limits to 2 and 4 times the standard level. Every track made in the app carries SynthID, Google's inaudible watermark for identifying content made with its AI.
How to make a song with AI, step by step
These steps apply to any generator that takes lyrics and a style description. The time goes into steps 2 and 3, not the generating.
- Decide what the song is for. A 30-second clip for a video, a full song to play at a party, or music under a slideshow. That decides the length and whether you need vocals.
- Collect the details. Names, a place, a habit, one small event. "Grandpa Ray" is a start. "Grandpa Ray, who fixes every bike on Elm Court" is a verse.
- Write or generate the lyrics. Label each part [Verse 1], [Chorus], [Verse 2] so the tool knows what repeats. Keep verses to four lines.
- Write one style line. Genre and era, tempo, two or three instruments, and the voice: "Upbeat 90s pop rock, bright electric guitars, handclaps, energetic female vocals."
- Generate, then listen to the whole track. Check every word, especially names. If a name comes out wrong, spell it the way it sounds and generate again.
- Download and trim. Cut a long song down to the part you need in any audio editor, and fade out the last two seconds.
A short lyric sheet to test a generator with:
[Verse 1]
Ray's got a wrench in his back pocket
A bike on the porch with a squeaky chain
Every kid on Elm Court knows where to find him
Saturday morning, sun or rain
[Chorus]
Grandpa Ray, Grandpa Ray
Fix it up and ride away
Oil on his hands at the end of the day
Everybody knows Grandpa Ray
[Verse 2]
He taught Maya how to patch a tire
He taught Leo how to ride with no hands
Now the whole street rolls past his garage
Ringing their bells like a marching bandIf you'd rather write your own lyrics from the start, the guide to turning your lyrics into a song covers section tags, line lengths and how to make words fit a tune.
How to make a song with AI for free
Free routes exist, and each one comes with a condition you should know before you spend an evening on a song. As of September 2026:
| Option | Free allowance | Condition |
|---|---|---|
| Gemini app | Music generation with no paid Google AI plan, at standard limits | Adults only, with a 5-hour usage cycle and a weekly cap |
| ElevenLabs Music | Free plan, listed at about 3 to 5 tracks a month | Personal use only |
| Udio | 10 credits a day, then a 100-credit monthly limit, with no more than three 130-second songs a day | You can't download the audio, video or stems |
| Toybox AI | 20 credits once, when you sign up | Music Generator currently costs 95 credits per run and needs a Pro plan, so a song isn't covered |
For a one-off personal song, such as a birthday message, a free tier can be enough. For anything you plan to post, sell or play at a business, read the tool's terms first. The AI music generator roundup compares paid plans, downloads and commercial rights in one table.
Where AI songs fall short
Plan around these weak spots before you share an AI song or build a video around one:
- Words can come out wrong. A generator can slur a name, stress the wrong syllable or skip a line. Listen to every version before you share it.
- Length is loosely controlled. Gemini offers short or full. Toybox's Music Generator has no length setting at all. Trim the file when you need an exact length.
- Your own tune can't go in everywhere. Music Generator takes no audio uploads, so it can't follow a melody you hum. ElevenLabs accepts a voice recording or a loop as a starting point, and Udio's paid plans can generate from audio files you upload.
- Other people's voices are off limits. Spotify says vocal impersonation is only allowed on its service when the artist being imitated has authorized it.
- Downloads and rights vary by plan. A song you can't download or can't use commercially may still be fine for a group chat. Check the plan before you start editing.
Who owns a song AI made for you
In the US, the answer depends on what you wrote yourself. This is general information, not legal advice.
- Your lyrics are yours. Circular 1 from the U.S. Copyright Office says copyright exists automatically from the moment an original work is fixed, for example typed into a document.
- The AI's contribution may not be protected. On January 29, 2025, the Copyright Office concluded that AI output can be protected only where a human author determined enough of its expressive elements, and that writing prompts alone doesn't count. Human-written material that is perceptible in the output, such as your lyrics, can still be protected.
- The tool's terms decide what you may do. Toybox's Terms of Service say you own the rights to what you generate, subject to the Terms, and the kinds of output they name are images, text, video and code. Songs aren't named, so read the Terms and each platform's rules before you release or monetize one.
Posting on YouTube adds its own disclosure and Content ID questions, covered in the guide to making AI music for YouTube.
Make a song with vocals in Music Generator
Toybox's Music Generator suits songs about specific people and moments: you give it a title, a style and your lyrics, and it returns two versions to choose from. It needs a Pro plan and currently costs 95 credits per run. Pro currently costs $9.99 a month and includes 1,000 credits, which covers 10 runs (see plan prices).
- Name the song in "Song title" (80 characters at most).
- In "Describe your song", say how it should sound: genre, tempo, instruments, voice (up to 1,000 characters).
- With "Song type" set to "With vocals", paste your words into "Lyrics", or leave that box blank to get lyrics written for you.
- Under "More options", set "Voice" to Any, Female or Male.
- Press "Create song" and wait a few minutes for two versions. A run that fails gives its credits back.
- Play both, then download the MP3 you prefer or keep it in your gallery, since a song you don't save is gone after 24 hours.
Read the lyric sheet once more before you press the button, because a typo gets sung exactly as typed. For a gift with a name in it, the page on making an AI birthday song with their name has lyric templates by age and a pronunciation fix.