On this page
Short answer
Make short Seedance clips for each part of your song in KreatorFlow's video studio, attaching that part of the track as an audio reference and keeping one style line in every prompt. For the singing, upload a photo and your vocal to Lip Sync (4 credits per second at 720p, up to 5 minutes). Then put the song and the clips on the editor's timeline, cut on the beat and export.
Try this in the studio
Make an AI music video in 6 steps
- 1
Prepare the song and a plan
Export your finished song as an MP3 or WAV. If you will have a character sing, also export a vocal-only version (a stem from your project or your producer). Then write a one-line plan per section: intro, verse, chorus, bridge, outro, with the time each one starts. Seedance clips run 4 to 30 seconds, so each section becomes one or more clips.
- 2
Lock the look before you make any clips
Decide on one visual style and write it as a single line you paste at the start of every prompt, for example "Grainy 16mm film, warm tungsten light, shallow depth of field." If the video has a singer or a recurring character, save them once in Characters and mention their @handle in each prompt, so the face stays the same from clip to clip.
- 3
Generate the visuals in the video studio
Open the video studio, pick Seedance 2.5 or Seedance 2.0 and the shape you will post in. For a clip that should follow the music, attach that part of the song as an audio reference (up to 3 audio references, 2 to 15 seconds of audio in total on Seedance 2.0, and up to 30 seconds on Seedance 2.5) and describe what happens on the beat in the prompt: "the camera pushes in on the kick drum", "she spins on the drop". Make each section's clips, and keep the ones that fit.
- 4
Add the singing with Lip Sync
Open Lip Sync, add a photo or a saved character with a clear, uncovered mouth, and upload the vocal as an MP3 or WAV of 4 seconds to 5 minutes. Write "sings" in the optional direction, for example "Slow push-in. She sings with her eyes closed, then looks into the camera when the chorus starts." The video is exactly as long as the audio. It costs 4 credits per second at 720p and 6 at 1080p.
- 5
Lay everything over the song in the editor
Open the editor and import your song (MP3, WAV, M4A, AAC or OGG up to 15 MB) and your clips. Put the song on an audio track, then drag the clips onto the timeline above it and trim or split each one so the cuts land where the music changes. Line the Lip Sync clips up with the vocal they were made from, so the mouth matches the song. If a clip came with its own generated sound, make sure your song is the only thing you hear before you export. You can also make extra clips right in the editor's Generate panel with Sound switched off.
- 6
Add lyrics and export
For lyric text, type it as text clips or import an SRT or VTT file of your lyrics. Then export in the shape you need (9:16, 16:9, 1:1, 4:5, 4:3). A free account exports up to 1 min per video, so a full song needs Creator (up to 10 min) or above.
What the pieces cost
Studio credits from the same code that charges you. The price for your exact settings shows before you generate.
| Piece | Price | Example |
|---|---|---|
| Seedance 2.5 clip, 720p | 24 credits for 5 seconds | 140 credits for 30 seconds |
| Seedance 2.5 clip, 1080p | 50 credits for 5 seconds | 300 credits for 30 seconds |
| Seedance 2.0 clip, 720p | 15 credits for 5 seconds | 45 credits for 15 seconds |
| Seedance 2.0 clip, 1080p | 30 credits for 5 seconds | 90 credits for 15 seconds |
| Lip Sync (OmniHuman 2.0), 720p | 4 credits per second of audio | 720 credits for a 3-minute song |
| Lip Sync (OmniHuman 2.0), 1080p | 6 credits per second of audio | 1,080 credits for a 3-minute song |
| Editor, free account | 5 exports a month, up to 1 min each, 1080p | Small "Made on KreatorFlow.ai" mark |
| Editor, Creator | 60 export minutes a month, up to 10 min each, 2K | No mark |
Example: 12 Seedance 2.0 clips of 15 seconds at 720p cover a 3-minute song for 540 credits, and a 30-second sung chorus adds 120 credits at 720p. 1080p video clips are on Pro, Max and Business. Lip Sync needs a paid plan (from $15 for the first month on Creator) or a credit pack (from $8). Credits for a generation that fails come back automatically.
The three parts of an AI music video
Almost every AI music video is built from the same three pieces. The first is the visuals: short generated clips of scenes, places, moods and abstract shots that carry the song. On KreatorFlow these come from Seedance, ByteDance's video model, which runs on BytePlus, a ByteDance company (KreatorFlow is an official partner of BytePlus, a ByteDance company).
The second is the performance: a singer whose mouth moves with your vocal. That is Lip Sync, which runs OmniHuman 2.0, an audio-driven model that animates a photo to match the voice track, including long notes and fast rap. The third is the edit: the song on a timeline, with the clips cut to its sections, and the finished file exported.
You do not need all three. A visualiser or lyric video can be visuals and text only. A performance video can be one long singing take with a few cutaways. Decide which kind you are making before you spend credits.
Making clips follow the beat
There are two ways to get visuals that feel tied to the music. The first happens at generation: attach the part of the song a clip will sit under as an audio reference, and say in the prompt what should happen on the beat. The model hears the track while it makes the clip. Audio references can be 2 to 15 seconds of audio in total on Seedance 2.0, and up to 30 seconds on Seedance 2.5, so trim the song into short sections before you attach them.
The second happens in the edit, and it matters more. Cut on the beat: put a cut on the first beat of a bar, on a snare hit or on the drop, and the video will feel in time even if the clips themselves are calm. Fast songs suit short clips and many cuts; slow songs suit long clips and slow camera moves.
A good habit is to generate a little longer than you need. A clip that runs past the beat can always be trimmed in the editor, but a clip that ends too early leaves a gap.
- Prompt one action per clip and name when it happens: "on the drop", "as the chorus starts", "on every kick".
- Match camera speed to tempo: slow push-ins for ballads, handheld and whip pans for fast tracks.
- Keep a few abstract clips (light leaks, smoke, city lights) to cover instrumental gaps.
- Generate intro and outro shots last, once you know how the middle looks.
Keeping one look across every clip
A music video falls apart when every clip looks like a different film. Three habits keep it together. Paste the same style line at the start of every prompt (film stock, lighting, colour palette, lens). Save your singer or main character in Characters and use their @handle in every prompt, instead of describing them again each time. And keep the shape the same: pick one aspect ratio in the studio and do not change it halfway.
If you have artwork for the release, such as the single cover, add it as an image reference so the colours and mood carry over. The video studio takes up to 9 image references per clip.
Example clip prompt with a style line
Grainy 16mm film, warm tungsten light, shallow depth of field. @nova walks down an empty neon street at night, rain on the ground. The camera tracks beside her, then pushes in on her face as the chorus starts.
Getting the singing right
Lip Sync takes a vocal track of 4 seconds to 5 minutes, so most songs fit in one continuous take. The vocal matters more than the prompt: a vocal-only stem gives the most accurate mouth movement. A full mix also works if the voice sits well above the music, but drums and loud instruments can blur the timing, especially on fast verses.
KreatorFlow does not split vocals out of a finished mix, so ask your producer for the stem or export it from your own project. Trim long instrumental intros before you upload: the singer would just sit silent through them, and you pay per started second of audio. Test the hardest part, usually the chorus, at 720p before you render the whole song at 1080p.
The photo needs one clear face with the mouth uncovered (no microphone in front of the lips), in a head-and-shoulders or half-body shot. Singers, drawn characters, anime characters and animals all work.
What KreatorFlow can't do for a music video
It helps to know the limits before you start, so you can plan around them.
- No single clip covers a whole song. Seedance clips are 4 to 30 seconds, so the visuals are always several clips joined in the editor.
- The editor does not detect beats or snap cuts to the beat. You place the cuts yourself by ear and by the waveform.
- There is no vocal separation. Bring a vocal stem if you want the cleanest lip sync.
- Auto captions turn speech into text and may struggle with sung lyrics. For lyric videos, type the lines or import an SRT or VTT file.
- The editor is built for a laptop or desktop browser, not a phone.
- Lip Sync needs a paid plan or a credit pack, and one take tops out at 5 minutes. Split a longer track and join the takes in the editor.
- A generated clip will not exactly match a shot list. Expect to make a few versions of the important shots and keep the best one.
Your song, your rights
You can only use a song you have the rights to: your own recording, an original you made, or a track licensed for this use. KreatorFlow does not clear music for you. Posting a commercial song you do not own can get the video muted or removed on social platforms and can infringe the rights holder's copyright.
The same goes for faces and voices. Only make a real person sing if you are that person or have their permission, and never use it to make someone appear to perform something they did not. When you share a realistic singing video of a real person, say that it was made with AI.
Frequently asked questions
How do I make an AI music video for my own song?
Generate short clips for each section of the song in the video studio (attach part of the song as an audio reference so the clip follows it), make the singing parts with Lip Sync from a photo and your vocal, then import the song and the clips into the editor, cut the clips to the music and export.
Can AI make a music video that matches the beat?
Partly. You can attach a section of the song (2 to 15 seconds of audio in total on Seedance 2.0, and up to 30 seconds on Seedance 2.5) as an audio reference and say in the prompt what happens on the beat. The tight sync comes from the edit: you cut the clips on the beat in the editor, which does not detect beats for you.
Can I make an AI character sing my song?
Yes. Upload a photo or use a saved character in Lip Sync, add your vocal as an MP3 or WAV of 4 seconds to 5 minutes, and write "sings" in the direction. It costs 4 credits per second of audio at 720p.
How much does an AI music video cost?
It depends on how much you generate. Covering a 3-minute song with 12 back-to-back Seedance 2.0 clips of 15 seconds at 720p is 540 credits. A singing take of the whole song at 720p is 720 credits. Most videos mix the two and reuse clips, so they cost less. Adding the song in the editor is free.
How long can an AI music video be?
As long as your edit. Each Seedance clip is 4 to 30 seconds and each Lip Sync take up to 5 minutes, and the editor joins them. One export can be up to 1 min on a free account, 10 min on Creator and 30 min on Pro.
Can I use a famous song?
Only if you have the rights to it. Use your own music or a licensed track. A song you do not own can get the video muted or taken down, and KreatorFlow does not license music for you.
