Make or bring a song, build a still for the chorus, press Animate it, then repeat for the verse and the drop and join the clips to the track.
This session
My creationsYour results appear here. Every result is also saved in My creations; if a run fails, the credits go back to your balance.
Two vertical clips from original tracks, a synth-pop chorus and a lo-fi verse, remixed for this page from 10-second lovart.me originals, with the four section stills beside them. Chorus cut, about 8 seconds at about 122 BPM: two seconds close on the singer at a vintage mic, then a cut to her profile as the camera pulls back to the whole rooftop at golden hour. Verse loop, about 9.5 seconds at about 77 BPM: a musician playing a small keyboard, the camera drifting to the rainy window and the desk lamp and back to him. Both have sound; turn it on in the player.
Remixed from a lovart.me clip: Seedance 2.0 Fast, 10 s, 9:16, 720p, with a 12-second clip of an original synth-pop chorus with a wordless la-la hook (Suno V6) as the audio reference; the sound was generated by the model from that reference · re-cut on the beat, reframed, colour graded · no new AI generation · made 2026-11-05
Remixed from a lovart.me clip: Seedance 2.0 Fast, 10 s, 9:16, 720p, with a 12-second clip of an original lo-fi instrumental (Suno V6) as the audio reference; the sound was generated by the model from that reference · re-cut on the beat, reframed, colour graded · no new AI generation · made 2026-11-05
Still: GPT Image 2.5 Flare, image to image, 9:16 · still only · made 2026-11-04
Still: GPT Image 2.5 Flare, image to image, 9:16 · still only · made 2026-11-04
Still: GPT Image 2.5 Flare, image to image, 9:16 · still only · made 2026-11-04
Still: GPT Image 2.5 Flare, image to image, 9:16 · still only · made 2026-11-04
Both singers are original adult characters, not real artists, and both songs are original. The two clips were made once on lovart.me with Seedance 2.0 Fast, each with a 12-second clip of a Suno V6 song attached as an audio reference, so the sound you hear was generated by the model from that reference rather than taken from the song file. For this page they were re-cut on the beat, reframed and colour graded, with no new AI generation. Section stills were made with GPT Image 2.5 Flare on November 4, 2026. Last checked .
Step one: write the track on AI Music or bring your own; step two: build a still for each section, animate it, and join the clips to the song in any editor.
Describe the song on AI Music in one line, or upload your own. Note the BPM and the second where the chorus starts.
Upload one singer photo, pick Chorus, Intro, Verse or Drop, and press Build the still. Keep the same photo for every section.
Press Animate it for each section, download the clips and line them up on the song in any editor, cutting on bar lines.
Mark intro, verse, chorus and drop, give each one still, and set clip length by bars: at 118 BPM, eight bars is about 16 seconds, so make two 8-second clips.
Usually a short, smaller or watermarked clip: HeyGen and Freebeat start free with limits, and here the still step and AI Music let you test the idea before a video run.
Write who, where, the motion on the beat and the camera move, and end each prompt on the frame the next section starts from.
Keep the same person from my photo: same face, hair and clothes. They sing full voice into a vintage handheld microphone on a city rooftop at golden hour, one arm raised, wind in the hair, string lights overhead, low angle, shallow depth of field. Vertical 9:16 music video frame, the singer fully in view. No text, no lyrics on screen, no logos, no real celebrities. Use the photo as the first frame. One continuous shot, no cuts. Arm swings down on the downbeat, a step toward the camera every two beats, hair moving in the wind, the camera rises slowly from a low angle. Keep the same face, hair and clothes in every frame. No text, no logos.
Keep the same person from my photo: same face, hair and clothes. They sit on a low brick parapet on a city rooftop in late afternoon light, singing softly into a vintage handheld microphone, medium shot at eye level, water towers behind. Vertical 9:16 music video frame, the singer fully in view. No text, no lyrics on screen, no logos, no real celebrities. Use the photo as the first frame. One continuous shot, no cuts. Small head nods on every beat, the free hand taps the knee, the camera drifts slowly to the side. Keep the same face, hair and clothes in every frame. No text, no logos.
Not in effect mode, which animates frames without your audio; for motion that follows the song, attach it as an audio reference on the Seedance 2.5 model page, and keep close-up singing short.
Export 9:16 for TikTok and Reels, add YouTube's AI label when the music or scene looks real, and trim a 3 to 8 second loop without singing for Spotify Canvas.
You can start free: HeyGen and Freebeat have free tiers with limits on length, resolution or downloads. Here you can check each section still before any video run, and the cost shows on each button.
Yes, or make one: describe the track on AI Music, download it, and note the BPM and where the chorus starts. Those two numbers set every clip length.
Count bars: at 118 BPM eight bars last about 16 seconds, so a three-minute song is roughly 12 to 20 clips joined in an editor.
Yes. Build the section clips, then add timed lyric captions in your editor. Keep words off the generated frames, because image and video models misspell them.
No. Songs made on Suno's free plan are for non-commercial use and Suno keeps ownership; songs made during a Pro or Premier subscription carry commercial use rights. Lyrics you wrote stay yours.
Each clip is a new run. Use the same still as the first frame of every section and keep the singer description word for word.
Write the track first, or put the singer on a fixed stage.