To make music with AI in 2026, you write lyrics, a short style description and a song structure in a service such as Suno, and the model generates a full track with vocals and instruments in a few minutes. The hard parts are elsewhere: good lyrics, pronunciation in less common languages, and rights. Songs made on a free plan usually cannot be used commercially.
A lot changed in 2025 and 2026: the major labels first sued the AI music services, then licensed some of them, which changed what you can download and where you may use it.
Below is the workflow step by step, followed by what to check before you put an AI song into an ad, a video or a YouTube upload.
Which AI tool should you use to make a song?
Three services come up most often: Suno, Udio and Eleven Music from ElevenLabs. All three turn text into music; the real difference is in the terms. The table reflects October 2026, and plans change often, so check the official pages.
| Service | Strong at | Free plan | Commercial use |
|---|---|---|---|
| Suno | Full songs with vocals, section editing, stems | 50 credits a day, no downloads | Paid plans only (Pro, Premier) |
| Udio | Vocals and arrangement | Listen and share on the platform | Downloads switched off in October 2025 |
| Eleven Music | Instrumentals and songs, regenerating one section | Personal use only | Paid plans, with exclusions (film, TV) |
In September 2026 Suno released a new model generation: v6, v6-wild and v6-mini, which the company says were developed with Warner Music Group, BMG and Believe; its older models are being retired. v6 and v6-wild are for Pro and Premier subscribers; v6-mini is available to everyone. As of October 2026, Pro includes 20 song downloads a month and Premier 60, with extra downloads sold separately.
ElevenLabs says Eleven Music was built in partnership with artists, labels and publishers and is cleared for broad commercial use. Rights depend on the plan: the free plan is for personal use only, and the paid self-serve plans exclude film, TV and major studio games.
How to make a song with AI: 8 steps
We use Suno as the example because every stage of making a song is in one place there: lyrics, style, editing and stems. The same logic works in other services.
- One-sentence idea. Who the song is for, what mood it needs and where it will play: a 15-second reel, a birthday surprise or a café jingle.
- Lyrics. Write them yourself or with an AI assistant, but edit the final version by hand. Lyrics you write are your work and the best-protected part of the song. For how the main assistants handle writing in Georgian and other languages, see ChatGPT, Claude or Gemini: which for what.
- Structure tags. Mark sections on their own line in square brackets: [Intro], [Verse], [Chorus], [Bridge], [Outro]. This is a widely used convention among Suno users. The model often follows it, but there is no guarantee.
- Style prompt. Four to eight words or phrases: genre, tempo (BPM), instruments, vocal type and mood. Write the style in English even when the lyrics are in another language, because genre names are less ambiguous that way. Describe the sound in words instead of naming a famous artist.
- Generate and choose. Each request gives you several versions. Listen on headphones and pick the best melody and chorus, not the best mix; the mix can be fixed later.
- Edit. Replace a weak section, extend a short track, or make a version in another style. In Suno’s v6 generation you can change one section in plain language while the rest of the song stays the same.
- Stems. A stem is a separate track of the song: vocals, drums, bass. Suno Pro offers two stem separation modes and Premier three. With stems you can mix in a free editor such as Audacity or in your video editor.
- Download and keep a record. On a paid plan, download through the official button: under Suno’s terms, only tracks downloaded through an approved channel may be used commercially. Save the lyrics, style prompt, date and plan name.
Here is a complete request you can copy and adapt. The style line lists genre and instruments, the sections are in brackets and the chorus repeats:
Style: indie pop, 100 BPM, acoustic guitar, soft drums, warm female vocal, summer morning mood
[Intro]
[Verse]
Morning in the park, the sun is rising slow,
coffee in the air and nowhere I need to go,
the bus is running late but I don't really mind,
today the good things come in their own time.
[Chorus]
Tbilisi morning, warm and bright,
I hear the city singing in the light,
Tbilisi morning, loud and new,
today it all begins with you.
[Outro]
It all begins with you.The general rules of prompting (role, context, example, format) apply to music too. We cover them in detail in our guide on how to write a good prompt.
How to write lyrics in Georgian or another less common language
Languages like Georgian appear far less often online than English, so models do not always get pronunciation and stress right. You can still make a good song if you prepare the lyrics for the model:
- Short lines. Six to ten syllables per line, with lines of similar length within a section.
- Simple words. Long consonant clusters are hard to sing. Choose a synonym when a word keeps going wrong.
- Numbers as words. Write “twenty twenty-six”, not “2026”.
- Repetition in the chorus. One clear phrase repeated two or three times helps both the model and the listener.
- Chorus first. Test only the chorus in a short version before generating the whole song, and you will save credits.
- If a word still sounds wrong, split it with hyphens or change the word order.
If the vocal still sounds foreign, keep the AI instrumental stem and record the vocal yourself. That also increases the human contribution, which matters for rights.
Change one thing at a time. If you change the style, the lyrics and the structure at once, you will not know what improved the result. Keep a simple log of every version: the style line, what you changed and what you liked. After ten generations you will have your own working template.
Who owns a song made with AI?
Your lyrics are yours; rights in the AI-generated music depend on the service’s terms and national law. Suno’s Terms of Service (last revised 10 August 2026, in force from 3 September) say:
- Paid plans. Suno assigns its rights in the output to you, subject to its terms. Only tracks downloaded through an approved channel may be used commercially, and remixes may not.
- Free plan. Output is for personal, non-commercial use only.
- No guarantee. Suno does not promise that any copyright will exist in the output, and similar output may be generated for other users.
- Labelling. You may not remove the watermark or metadata Suno adds to a track.
In its January 2025 report, the U.S. Copyright Office concluded that merely providing prompts is not enough for authorship. Human creative contribution is protected: your own lyrics, selection, arrangement and modifications. In practice, the more you contribute, the more you own. Rules differ between countries, so talk to a lawyer before a large commercial project.
What is happening between Suno, Udio and the labels?
The major labels sued Suno and Udio, arguing that the models had been trained on their recordings without permission. Since then:
- October 2025. Universal Music Group and Udio settled and agreed to build a new licensed platform. Udio switched off downloads immediately, so songs stay inside the platform. The new licensed platform was planned for 2026, so check Udio’s current terms on its site.
- November 2025. Warner Music Group settled with Suno and signed a licensing deal. Suno committed to licensed models, download limits and retiring its old models.
- 31 July 2026. The Munich Regional Court ruled in a case brought by GEMA, the German music rights society, that Suno infringed the copyright in six well-known songs. The judgment is not yet final, and Suno says it is exploring all options, including an appeal.
- September 2026. Suno released the v6 models, developed with Warner, BMG and Believe. Universal and Sony are not part of that deal.
The rules may change again, so keep a record of the plan and date for every song, along with the downloaded file.
How to use AI music in videos, ads and on YouTube
The most common use is background music for a reel, a product ad or a logo animation. Here is what each case needs.
| Where you use it | What you need |
|---|---|
| Personal video for friends | A free plan can be enough if sharing on the platform works for you |
| Business reel or ad | A paid plan and an officially downloaded file |
| Monetised YouTube channel | A paid plan and the AI disclosure when uploading |
| Client project | A paid plan, records and a written agreement with the client |
| Film or TV | A separate licence check: some plans exclude this outright |
YouTube requires creators to disclose realistic AI content at upload, and its examples explicitly include AI-generated music. Creators who consistently fail to disclose may have content removed or face suspension from the Partner Program. Also, do not imitate a real singer: Suno’s terms forbid presenting a track as if it were created by someone who did not create it.
Music is half of a video; for the shots, animation and voice-over, read AI video and animation.
Course at TDSAI for Motion Designers: Video, Animation and Sound4 weeks · 12 hDetailsFrequently asked questions
Can I make a song with AI for free?
Yes. Suno’s free plan gives you 50 credits a day with the v6-mini model, but as of 2026 free-plan songs cannot be downloaded or used commercially; you can only play and share them on the platform.
Can Suno sing in languages other than English?
Yes, you can enter lyrics in other languages and the model will sing them, but pronunciation and stress are not always accurate in less common languages such as Georgian. Short lines, simple words and numbers written as words help. If the vocal is not good enough, keep the AI instrumental and record the vocal yourself.
Can I upload an AI song to Spotify?
It depends on the service’s terms and on your distributor. Suno says songs made with its new models will also become eligible for distribution through Believe and TuneCore, and commercial use requires a paid plan and an official download. Udio downloads have been switched off since October 2025. Read your distributor’s AI rules before uploading.
Who owns a song made with AI?
Lyrics you write are yours. Suno assigns its rights in the AI-generated part to paid users but does not promise that copyright exists at all. The U.S. Copyright Office holds that prompts alone are not authorship, while human creative contribution is protected.
Which AI is best for music?
It depends on the task. For a full song with vocals and editing, Suno keeps every stage in one place. For instrumental background music in ads, Eleven Music’s commercial terms are convenient. Udio currently suits only work that stays inside its platform.
Where to start
Start with one short jingle: write a four-line chorus, generate three versions and pick the best. If you need music for ads or videos and want to learn the whole process from script to finished spot, see the course AI for Motion Designers: Video, Animation and Sound in Tbilisi. If you are not sure which direction suits you, contact us for a free consultation. Courses for business, design and teaching are on our AI courses page.
Sources
- Suno Pricing, Suno
- Terms of Service, Suno
- Introducing v6, Suno
- Udio disables downloads after UMG deal, Digital Music News
- Munich Regional Court stops Suno using GEMA-protected music, JUVE Patent
- Copyright Office Releases Part 2 of AI Report: Copyrightability, U.S. Copyright Office
- Eleven Music, ElevenLabs
- Disclosing use of altered or synthetic content, YouTube Help
