Can AI make a new song that sounds like me singing?
Yes. AI can write and sing a brand-new song in a voice cloned from a short sample of yours, with no existing track needed. On Floniks, open AI Cover, choose "Sing a new song", and add a 10–30 second sample of you talking normally. Then paste your own lyrics, or describe the song and let it write them, pick a style, and create. It usually takes about two minutes and can return more than one version to choose from.
How "Sing a new song" differs from a cover
Floniks AI Cover has two tabs that solve different problems. "Cover a song" starts from a track you upload: it keeps the melody, lyrics, timing and instrumental, and replaces only the lead vocal with your voice. "Sing a new song" starts from nothing but your voice. A voice is cloned from your sample, the lyrics come from you or are written from your description, and a new track in the style you ask for is generated and sung in that voice. That makes it the right choice when you have no song file at all: a birthday song for a friend, a jingle, a demo of lyrics you wrote, or simply hearing yourself on a genre you would never record. You can open this mode directly at /ai-cover?mode=voice_song.
What you need to provide
Only two things are required: a voice sample and either lyrics or a song idea. For the sample, 10–30 seconds of natural speech in a quiet room is the recommendation; the page accepts 8 to 60 seconds in MP3, WAV, M4A or WebM up to 20 MB, and you can record straight in the browser with "Record now". For the words, either paste lyrics (up to 5,000 characters, ideally with section markers like [Verse] and [Chorus]) or leave lyrics empty and describe the song, for example "a happy birthday song for my sister Mia who loves the ocean". Optional fields let you set a title, a style such as "acoustic pop, warm, guitar", and the vocal: match my voice, male or female.
Getting lyrics and style right
The more specific the input, the less generic the song. Structured lyrics with clear verse and chorus sections give the music model a shape to follow, while a description works best when it names concrete details: who the song is for, a place, an inside joke, the mood. Keep the style field short and concrete, naming a genre, a feel and one or two instruments, rather than a list of artists. If the first result is close but not right, change one thing at a time: tighten the chorus, swap the style, or re-record a cleaner voice sample. Because each generation is billed separately, small deliberate changes are cheaper than starting over from scratch.
Results, credits and reusing your voice
Generation runs in the background, so you can leave the page; finished songs appear under Creations, which are private by default, where you can play and download each version. Each new song costs a flat number of credits that the page shows before you create, and credits are refunded automatically if a generation fails. If you are signed in, tick "Save this voice so you can reuse it next time" when you add your sample. The saved voice then appears under "Use a saved voice" in both tabs, and once it has been used for a new song, later songs skip the voice-cloning step and finish sooner.
Related questions
Build it on Floniks
Image, video, digital humans, and reusable workflows on one canvas. No card required.
Explore Floniks