How to use AI Music Video Generator
Add the track and an artist photo
Under "Audio file (optional)", upload the part of the song the shot is for, ideally with clear singing. Use MP3, WAV or M4A, up to 30 MB. Under "Your image (optional)", add a sharp, front-facing photo of the artist so the same face carries across shots. Use JPG, PNG or WebP, up to 10 MB.
Describe the picture, not the lyrics
In "Describe the shot", write who is where, the light and the camera. For example: "singer in a rainy neon street, camera pushes in on the chorus, cut on every beat". Lyrics in the song are not read as instructions. A chip under "Style preset" fills the box with a complete music-video look.
Pick the model, length and resolution
Under "Model", only MiniMax H3 uses your audio file here. It syncs the shot to the track, at 2K for 5 to 15 seconds. Seedance 2.5 makes up to 30 seconds at 480p or 720p without the track. Cut that clip to the music later. Set "Duration" and "Resolution" within those limits.
Check the sync, then build the video
Press "Generate to the track" and check that cuts land on the beat and lips follow the vocal. If the lips drift, use a part with clear singing and a front-facing photo of the artist. A longer video is several shots: keep the best takes and cut them together.
What a real job costs here
The cheapest model on this page charges 17 credits per second and the most expensive 27 credits per second. At the best pack rate that is about $0.1809 up to $0.2874 per second. A 6-second clip costs 102 credits, roughly $1.09.
| Pack | Credits | Seconds with MiniMax H3 |
|---|---|---|
| $9 | 750 | 44 |
| $19 | 1,650 | 97 |
| $39 | 3,500 | 205 |
| $79 | 7,300 | 429 |
| $149 | 14,000 | 823 |
Choosing between the models on this page
MiniMax H3: 2K video with native stereo audio and audio references. Native 2K at 24fps with stereo sound in one pass. Watch out for: 2K only at launch, 768p tier in closed beta.
Seedance 2.5: Long single takes with dialogue, and reference-heavy work built from files you supply. Native 30-second generation in one pass. Watch out for: no 1080p or 4K on this route; 480p and 720p only.
The track is the input now
MiniMax H3 is the default engine on this page, and it is the first one here that listens. Upload the song, or the eight bars you are shooting for, and the model reads tempo, drops and vocal. The cut follows the music instead of a written guess about it.
What syncs and what does not
Cuts land on the beat, camera moves ride the tempo and a singer's lips follow the vocal. That is what the audio reference does well. It does not read lyrics as instructions. Write the visual you want and let the sound set the timing.
Build it one shot at a time
A music video is a stack of shots, not one render. Generate five to fifteen seconds per shot from the same section of the track. Keep the winners and cut them together. Each shot is billed per second, so drafts stay cheap.
Keep the face and the look
Add a sharp still of the artist under "Your image (optional)" and the same face carries across shots. Use the same still for every shot to hold the look. Each run takes one image and one audio file.
Seedance stays as the second engine
Seedance 2.5 does not use your audio file on this page. Pick it for a longer clip from the description alone, up to 30 seconds at 480p or 720p. You will cut it to the music later. It costs 27 credits per second at 720p; MiniMax H3 costs 17 credits per second at 2K.
Frequently asked questions
Can I upload my own song?
Yes. Attach the audio file and MiniMax H3 cuts the shot to it. Lips follow the vocal and edits land on the drums. Seedance ignores the file.
Does it sync lips to the vocal?
Yes, on the vocal in the audio you attach. Give it a front-facing artist and a section with clear singing, and describe the performance in the prompt.
Can it generate the song too?
Not on this page. Make the track in the AI music generator first, then bring the file here as the reference.
How long can one shot be?
Five to fifteen seconds with MiniMax H3, at 2K with stereo sound. Seedance 2.5 goes up to 30 seconds at 480p or 720p, but without your track. Longer videos are several shots cut together.
Is there sound in the output?
Yes. H3 returns stereo audio in the same pass, and with a track attached the output follows it.
What does it cost?
17 credits per second with MiniMax H3 at 2K. Seedance 2.5 costs 27 credits per second at 720p and less at 480p. The exact total shows before you generate.
