How do you make a music video with AI?
By Alverson Layne, Founder
Founder of FXI Studio, building the AI film studio.
Published October 6, 2026 · Updated October 6, 2026
The short answer
Start with the song: generate one from your own lyrics or upload a track you own. Split it into sections, draw a still for every shot, then turn the stills into video: singing shots driven by short slices of the song, and chained camera moves for the instrumental parts. Cut everything to the full track and add lyric captions.
TL;DR
Treat the song as the script. Every sung line becomes a singing shot that hears only its own slice of the audio, and every instrumental gap becomes a camera move. A 58-second cut planned this way costs about 356 credits in FXI Studio.
Start with 65 free credits. Creator plan $15/mo.
Try this prompt
“Music video shot. A singer in a flowing yellow dress twirls on a rooftop at golden hour as the city lights blink on behind her. Slow orbit, soft lens flare, warm dreamy grade.”
Opens the studio with this prompt filled in · 65 free credits
Start with the song, not the pictures
A music video is cut to its audio, so lock the song before you draw anything. You have two routes. Upload a track you wrote or have the rights to use, or generate one in Create.
To generate, pick MiniMax Music 2.6 (18 credits a track), leave Instrumental off, and put a style line and your lyrics in the same prompt: "Warm indie folk, fingerpicked acoustic guitar, soft brushed drums, upright bass, one clear female lead vocal, 96 BPM. Sing these exact lyrics:" followed by the lyric sheet. Listen to the whole take before you move on: a take can add or repeat a line, so keep the one that sings your words.
For an instrumental bed under a lyric-free video, turn Instrumental on, or try Lyria 3 Pro, which works from a prompt only and suits scores better than word-for-word vocals.
Map the song into shots
Write the song out as a shot list: one row per section with its start time. Sung lines become performance shots. Intros, instrumental breaks and the outro become camera moves through the world of the song.
A one-minute cut is roughly 4 sung shots of about 7 seconds and 6 camera moves of 5 seconds. A full three-minute song is the same plan three times over, which is why it pays to finish one chorus first and only then commit credits to the rest.
Pick one setting and one time of day per section. A pier at golden hour for the verses and the same pier at night for the chorus reads as one story; six unrelated locations reads as a slideshow.
Design one performer and keep them the same
Draw the singer once as a reference sheet on Nano Banana 2 (4 credits): front, three-quarter and profile, plain background, with two or three details a viewer remembers, like a long emerald dress and dark curly hair. Then save the performer as a character in your Cast in the Lab, so the same description goes into every prompt they appear in.
Make the performer original. Use a real person's face or voice only with their consent, as the acceptable use policy requires, and an original singer is yours to build a channel around.
Draw one still per shot with the sheet attached: the singer at the mic for each sung line, and a first and last key frame for each camera move. That is 11 stills for the cut on this page. Stills are where you fix framing and light cheaply, before any video credits.
Make the singer sing: one song slice per shot
Cut the song into one short audio piece per sung shot in any audio editor and export each as an MP3. Start each piece just before the first word and end it just after the last. Keep every piece under 15 seconds: that is the limit for an audio reference.
In Create, pick Seedance 2 Mini. Attach the shot still and the performer sheet as reference images, attach that line's audio piece in the audio slot, and set the length to the piece rounded up to whole seconds. Prompt the performance: "The singer from image 1 (design exactly as image 2) sings the vocal from audio 1 into the vintage microphone, mouth shapes matching every word, slow push-in, warm string lights, gentle sway on the beat."
Because each clip hears only its own words, the mouth shapes follow them. One rule to know: a clip with an audio reference cannot also take a first frame, so the still goes in as a reference instead. Seedance 2.0 Fast gives a closer, crisper face for three times the credits; save it for the shot people will pause on.
Chain camera moves for the instrumental parts
For the intro and the breaks, use P-Video with one key frame as the first frame and the next as the last frame, and describe only the move between them: "Slow, steady crane down from the string lights to the singer on the pier stage, ending exactly on the attached last frame. One unbroken camera move, no cut."
Chain the moves so each one starts on the frame the last one ended on. If a rendered clip lands somewhere slightly different from the plan, pull its real last frame and start the next shot from that. The result plays as one continuous camera, which is what makes a music video feel shot rather than assembled.
Cut to the full track and add lyric captions
Lay the full song on the timeline first, then place each singing clip where its slice starts, trimmed to the slice length. Use the full song as the soundtrack rather than each clip's own audio, so the music never jumps and the lip-sync stays lined up. Drop the camera moves into the gaps and speed them up a little if they run long.
The editor in the Lab can transcribe the song into word-timed captions and light up each word as it is sung, with your choice of font, size, colors and position. That turns the same cut into a lyric video for free.
What is free and what needs credits
New accounts start with 65 free credits and no credit card, about 16 images or 3 free videos (2 today, 1 more next week), plus 22 more credits each week until the free balance is back at 65. Free accounts make video with P-Video only, 5 seconds at 720p with no audio input, so the free credits cover the song, a few stills and one camera move: about 59 credits.
Singing shots run on Seedance, which needs any credit pack or plan. A full pass of the 58-second cut on this page is about 356 credits, inside the 600 monthly credits of The Creator plan ($15/mo).
| Step | What you make | Model | Credits (one pass) |
|---|---|---|---|
| 1. The song | One track from your own lyrics (0 if you upload a song you own) | MiniMax Music 2.6 | 18 |
| 2. Performer and shot stills | A performer sheet plus 11 shot stills at 4 credits each | Nano Banana 2 | 48 |
| 3. Slice the song | One short audio piece per sung shot | None (any audio editor) | 0 |
| 4. Singing shots | 4 shots of 7 seconds at 5 credits a second | Seedance 2 Mini | 140 |
| 5. Instrumental b-roll | 6 chained shots of 5 seconds at 25 credits each | P-Video | 150 |
| 6. Edit and captions | Cut to the full track, add word-highlight lyrics | Editor | 0 |
| Total | About 58 seconds of music video | 356 |
Estimate, not a quote. Prices are the published per-unit credit costs; stills are priced at the 1K size. Budget a retake or two on the singing shots, and more shots for a full-length song. On paid plans, Z-Image stills in Create use no credits, so b-roll key frames drawn on Z-Image cost 0 there.
How we tested this
The workflow is the one we ran end to end for our kids counting-song tutorial (song from typed lyrics, one consistent character, singing shots driven by song slices, a chained camera move, captions), applied to a music video. The budget below is a plan priced at FXI Studio's published per-model credit costs, not a finished film: 4 credits per Nano Banana 2 still, 5 credits a second for Seedance 2 Mini and 25 credits per 5-second P-Video shot.
Step by step
How to make a music video with AI
- 1Lock the song
Upload a track you own, or generate one in Create with MiniMax Music 2.6 from a style line and your exact lyrics. Cost: 0 or 18 credits.
Open Create - 2Map the shots
Write one row per song section with its start time: sung lines become singing shots, instrumental parts become camera moves. Cost: 0 credits.
- 3Design the performer
Draw one original singer as a reference sheet and save them to your Cast so the description goes into every prompt. Cost: 4 credits.
- 4Draw a still per shot
Same performer, same setting, the sheet attached to each still. Cost: 4 credits a still.
- 5Slice the song
Cut one audio piece per sung shot, each under 15 seconds. Cost: 0 credits.
- 6Make it sing
On Seedance 2 Mini, attach the still, the sheet and the line's audio piece, and prompt the performance. Cost: 5 credits a second.
- 7Chain the camera moves
Use P-Video from one key frame to the next for the intro and the breaks. Cost: 25 credits per 5-second shot.
- 8Edit and caption
Cut the clips onto the full song, add word-highlight lyric captions, and export. Cost: 0 credits.
Frequently asked questions
How do you make a music video with AI?
Lock the song first, by generating it from your lyrics or uploading a track you own. Draw a still for every shot, drive each singing shot with its own slice of the song so the singer lip-syncs that line, and chain camera moves for the instrumental parts. Cut it all to the full track. A 58-second cut costs about 356 credits in FXI Studio.
Is there a free AI music video maker?
You can start free in FXI Studio. New accounts start with 65 free credits and no credit card, about 16 images or 3 free videos (2 today, 1 more next week), plus 22 more credits each week until the free balance is back at 65. That covers the song, a few stills and a 5-second camera move. Singing shots with lip-sync need any credit pack or plan.
Can AI make a singer lip-sync to my song?
Yes. Give the video model a still of the singer plus a slice of the song as an audio reference, and prompt it to sing the vocal. Slice the song so each clip hears only one line, each piece under 15 seconds, and the mouth shapes follow those words. In FXI Studio this works on Seedance 2 Mini and Seedance 2.0 Fast.
Can I use my own song?
Yes. Attach your track, or slices of it, in the audio slot instead of generating one. Use music you wrote or have the rights to use, especially if the video will be posted publicly.
Can AI write and sing a song from my lyrics?
Yes. In Create, MiniMax Music 2.6 sings lyrics you type into the prompt alongside a style description, for 18 credits a track. Listen to each take and keep the one that sings your words in order.
How do I keep the singer looking the same in every shot?
Draw the singer once as a reference sheet with two or three memorable details, attach it to every still and singing shot, and describe them with identical words each time. Saving the singer to your Cast adds that description to every prompt automatically.
How long can an AI music video be?
As long as the song. Each singing shot covers one slice of up to 15 seconds and each camera move a few seconds, and you join them in the editor. Make one chorus first, check the lip-sync and the look, then build out the rest of the song.
Make your music video
Start with 65 free credits. Creator plan $15/mo.
Related guides
Hands-on tutorials
- TutorialHow to keep the same character across every shot
- TutorialHow to prompt P-Video for seamless transitions
- TutorialFXI SoundStage for Ableton Live: Install Guide + Session Workflow
Looking for a specific tool? Browse the documentation, the guides, or all tutorials.