Can CapCut's AI make music for your video, and when do you need a real song?

A bakery reel, 28 seconds. Steam off a tray, a hand dusting sugar, the door sign flipping to OPEN. You open the Audio tab in CapCut, pick AI music, type warm acoustic morning, light percussion, cheerful into the prompt box and press Generate. A minute later there are a few options in History, one of them sits nicely under the cuts, and you are done.

Then the owner watches it and says: "Lovely. Can it sing Rosa's opens at seven?"

That sentence is where this article splits in two. CapCut's AI can make music for a video, and for a lot of edits it is the shortest path you have. Whether it can carry a line of words that has to be heard, remembered and repeated across twenty videos is a different question, and so is whether the track is cleared for what the video is about to do. Below: what CapCut's AI music actually is according to CapCut's own pages, the licence each kind of CapCut music sits under, the jobs where CapCut is enough, and the jobs where you need a real song with lyrics, plus the route for making one.

A quick note on who this page is for. If you are choosing between free library tracks, platform commercial libraries and stock music for Shorts, Reels and TikTok, the comparison of licence tiers lives in royalty-free music for Shorts, Reels and TikTok, and that is the page to read. This one is for the reader who already edits in CapCut, found the AI music button, and wants to know how far it goes and when to step outside it.

sunomarket is an independent site. We are not affiliated with CapCut, ByteDance, TikTok or Suno, and nothing here is sponsored. Every CapCut claim below was checked on capcut.com at the time of writing; CapCut changes features by version and region, so confirm in your own build.

What CapCut's AI music actually does

CapCut has more than one music feature with "AI" in the name, and they do different things. Mixing them up is how people end up expecting a sung chorus from a tool that makes background beds.

AI music generator (text to music). CapCut's tool page describes it on the desktop editor: open the Audio tab, choose AI music, type into the "Describe your music" box and generate. The page gives prompt examples like energetic pop beat and cinematic ambient music, and says the AI picks rhythm, instruments and tempo from the words you use. Finished tracks land in a History list, where you preview them and add one to the timeline. CapCut also points to Auto mark beats, which puts beat markers on the track so you can time cuts to it. The page does not say whether these tracks contain vocals, and it does not state a length limit.

AI song maker. A separate CapCut page markets a song maker that "instantly" creates vocals, lyrics and instrumentals from a text prompt, with an AI lyric writer that drafts words from a topic and keywords, and a Voiceover feature for recording your own vocal on top. Exports are listed as MP3, WAV, AAC and FLAC.

AI singing voice generator (speech to song). In the CapCut app, CapCut describes a feature that turns a voiceover or a recorded clip into a song, adding melody and harmony, in styles it lists as folk, reggae, hip-hop, jazz and R&B. You speak, it sings back in a style. It is not described as singing typed lyrics.

Copyright check. CapCut's AI music page also mentions a built-in music copyright detector that looks for conflicts before you upload. Treat that as a warning light. It tells you about a possible match; it does not hand you a licence.

On price, CapCut's pages call the generator free and say you can generate and download as many tracks as you like. That is the vendor's own wording at the time of writing. CapCut runs free and paid tiers, and what sits behind which tier moves, so check inside the app rather than trusting a page.

Three kinds of CapCut music, three different licences

This is the part almost nobody reads, and it is the part that decides whether a client video gets claimed. "Music in CapCut" is not one thing. CapCut's own documents split it into categories with different rules.

Three details from those documents matter in practice.

First, the commercial library is fenced by platform. A track marked "commercial" inside CapCut is cleared for posting on CapCut, TikTok and TikTok for Business. The same video uploaded to YouTube, a client's website or a paid placement on another network sits outside that wording. This is the trap for anyone who cuts once and posts everywhere.

Second, the Materials License Agreement says nothing about AI music. It covers library music and platform materials. The AI generator's commercial promise lives on a marketing page, and the Terms of Service, which do bind you, stay silent on who owns AI output and add that some AI features are built on third-party AI services whose terms also apply. Nothing in that adds up to a scandal. It is simply a gap between what a tool page says and what a contract says, and gaps are where claims come from.

Third, United States users are under a separate agreement. The global text says use of CapCut in the US is governed by a CapCut US Materials License Agreement instead. If you are in the US, read the version your account is under, not a summary of the global one.

This is not legal advice, and the rules differ by country. For how rights in AI music are usually framed, read who owns AI-generated music, and for our own positions on what we generate, see the disclaimer.

A paper trail that costs five minutes

If you are going to rely on the "royalty-free" line for a client video, do what any careful editor does with a stock licence: make the evidence before you need it.

• Save the CapCut tool page with the date, or take a screenshot of the line about commercial use. Pages change; your memory of them does not count.
• Keep the exported audio file, not just the project. A track that only lives inside a project can vanish if the asset is pulled.
• Note whether the track carries any "commercial" mark in the app, and which platform the video is going to.
• If the client is a brand with lawyers, send them the Terms of Service link and let them decide. Their risk tolerance is their call, not yours.

None of this is paranoid. It is the same habit you would apply to a stock track you paid for, and a free AI track deserves the same care, even though it cost nothing.

What CapCut's AI music is genuinely good at

Take the position plainly: for a music bed under talking, CapCut's generator is the right tool, and paying another service to do the same job is money you do not need to spend.

Beds under a voiceover. A talking-head explainer, a product demo, a recipe. The music should sit under the voice and never compete with it. A prompt like soft lo-fi, mellow keys, no vocals, steady is all the brief it needs, and the result never leaves the editor.

Short lengths and fast turnarounds. A 15 to 60 second edit with a deadline. Generating inside the editor skips the download, import, trim and re-sync loop of an outside tool.

Cut-driven edits. Auto mark beats is a real advantage for montage work: you generate, mark beats, snap cuts. An outside track can be marked too, but having it in one place saves a round trip.

Many variations, low stakes. When you need ten different moods for ten throwaway clips, a free generator that lives in your editor wins on effort every time.

Where it is thin is the opposite case: a piece of music that has to be specific. A melody people can hum. Words they can quote. A sound that is the same in video one and video forty.

When you need a real song instead

A background bed and a song are different products. You need a song, with written lyrics and a sung vocal, in these situations.

The words are the message. "Rosa's opens at seven" is information. A brand line, a product name, a person's name in a birthday clip, a date in an event teaser. If the viewer has to hear and remember words, a mood track cannot deliver them. Prompted song makers can produce lyrics, but a line like that only lands if you control it word for word.

The song is the content. On TikTok, a sound can travel without the video. If you want people to reuse your audio, the hook has to be a line, not a texture. TikTok has an AI song feature of its own; what it does and when a full generator beats it is in how TikTok's AI song feature works, and getting your finished track listed as a reusable sound is covered in make your song a TikTok sound.

The series needs one identity. Twenty videos with twenty generated beds sound like twenty different shops. One song with a chorus you cut down to 6, 15 and 30 second versions sounds like one brand. If you are building a sonic logo or a jingle, the longer treatment is AI jingles and sonic branding.

The lyrics are in another language. A Russian, Spanish or Japanese lyric needs someone, or something, that writes in that language with stresses that fit the melody. Leaving it to a prompt is a lottery.

The video is a lyric video. CapCut's auto lyrics can caption a song on screen. Captions make weak lines more visible, not less. A lyric video needs lyrics worth reading.

And one case where you need neither: if the clip already exists and you just want music composed to its cuts, tools that take the video file as input are compared in AI that writes music for your video.

The route for a real song: words, then voice, then cut

A sung song for a video is three steps, and each step is a different tool doing what it is best at.

1. Write the lyric. This is where sunomarket fits, and only here. Our lyric generator takes a form brief and returns a title and a full lyric with [Verse] and [Chorus] section tags, ready to paste into Suno's Lyrics field. It does not sing and never returns audio. SM1 costs 20 credits ($0.20); SM2 and SM3 run an extra editing pass that flags weak lines and rewrites them, and cost 80 credits ($0.80). New accounts start at 0 credits, there is no free trial, and plans are on the pricing page.

2. Generate the audio in Suno. Suno is a different company; we do not resell it or speak for it. Paste the lyric into the Lyrics field and describe the sound in the style box. Rights depend on the plan: Suno's pricing page lists "No commercial rights" on the free plan and commercial use rights on Pro and Premier, and its help centre says that on a paid plan you own the song and keep commercial rights after cancelling, while on the free plan Suno retains ownership and you may use songs non-commercially. So for anything a client pays for, subscribe before you generate. The detail is in can you sell Suno songs.

3. Cut it in CapCut. Import the track. Now Auto mark beats, fades and the timeline work on a song you control, and the licence question is about your Suno plan, not about CapCut's library.

Here is the bakery spot as a brief and the shape of the lyric that comes back. The brief uses the generator's real form values; the lyric below is our own illustration of the format, trimmed to what a 30-second spot uses.

theme:     Rosa's bakery on Linden Street opens at seven; first tray of
           cardamom buns, a line of people in coats, the door sign flips
POV:       collective
language:  en
structure: loop
hook:      title-line
vocal:     female
emotion:   joy
age:       25-34, 35-44

Title: Open at Seven

[Verse]
Coats in a line on Linden Street
Fog on the window, the oven's heat
First tray out and the whole room turns
Somebody whispers: save me one

[Chorus]
We're open at seven, come as you are
Warm in your hands from the door to the car
We're open at seven, the buns are still hot
Rosa's on Linden, the best seat we've got

The title line opens the chorus twice, which is what the title-line hook asks for, and the collective "we" is the bakery speaking. For a 30-second cut you keep this verse and one chorus; for a 6-second bumper you keep "We're open at seven, come as you are" and nothing else. That last move is the reason to have words at all: the same line works at every length.

Two practical notes for the Suno step. Section tags are what keep a short song short, so tag the structure rather than hoping for length. And if you want the version under a voiceover too, generate a clean instrumental from the same style prompt so the bed and the song belong to one family.

Which tool for which video

Before you press Generate on the next client video

Ask one question first: does this video need to say something in words? If not, stay in CapCut, generate the bed, save the exported file and a dated screenshot of CapCut's commercial-use line, and move on. If yes, write the hook line before anything else, the one sentence the viewer must leave with, then build the song around it: brief it in the lyric generator or write it yourself, sing it in Suno on a paid plan, and bring it back into CapCut to cut. The line you write first is the one you will reuse in every length of every video after this one.