How to make an AI music video, from clearing the track to a beat map, shot list, generation, lip-sync and cutting to the beat, with published tool prices.
Method. Vendor docs and price lists (Google, Kling, sync.so, Runway, Suno), Adobe help and librosa documentation read November 18, 2026; beat tables and cost examples are our arithmetic, and no lip-sync tool was tested on singing. Vinc makes an AI film studio.
How to make an AI music video: the short answer
Start from the finished, cleared track. Map its beats, write a shot list that fits each phrase, generate silent shots, lip-sync the singing shots separately, and cut to beat markers in an editor.
The song has to drive the edit because no mainstream video model will follow it: Google's Veo 3.1 lists audio input as not supported, and Kling 4.0 says "uploading background music or sound effects is not supported".
The six steps in this guide. The song is cleared and beat-mapped before anything is generated, because no mainstream video model will follow a song by itself.
Step 1: Can you legally use the song in an AI music video?
A music video is a sync use, so sort out the rights before you generate a frame.
Someone else's recording, or your own released track. A recording carries two copyrights, the composition (publisher) and the master (label), so putting it to picture needs a sync licence and a master use licence. If you are the artist, check what your label and publishing deals already grant. Music Licensing for Film, Explained covers both.
A generated track. Terms depend on the plan. Suno's pricing page, checked 18 November 2026, lists Pro at $10 a month ($8 billed annually), Premier at $30 ($24) and Platinum at $120 ($100), and says you can "use your songs commercially on any paid plan"; the free plan has no downloads and no commercial rights. Google Cloud lists Lyria 3 Pro full-song generation at $0.08 per song on its Agent Platform pricing. AI Music for Film covers ownership and which generators exclude film use.
Keep the licence terms, plan and download date with the project; distributors and platforms can ask for them.
Step 2: How do you map the beats of a song?
A beat map is a list of timestamps for every beat, with markers for bars and song sections.
Automatically. The open-source Python library librosa has a beat tracker, librosa.beat.beat_track (documentation). It returns the estimated tempo in BPM and the beat positions, in seconds with units='time'. Check the result by ear: tempo changes, rubato and quiet intros can trip any beat tracker.
By hand in the editor. In Adobe Premiere, per Adobe's help page, you move the playhead and "Select Markers > Add Marker, or press M on the keyboard". Trailer editor Leonardo Oliveira, interviewed by No Film School (24 December 2025), marks every beat and colour-codes the first beat of every four bars: "This allows me to know exactly where I want to cut the music whenever I want to either extend it or shorten it."
The table is our arithmetic in 4/4. At 24 fps (Veo 3.1's output rate per Google's model page), only some tempos land beats on whole frames; at 100 BPM, cut points round to the nearest frame. See Frame Rate and Shutter Angle.
Step 3: How do you turn a song into a shot list?
Write one row per phrase or section, not one per idea. Useful columns: section (verse, chorus, bridge), start and end time, bars, shot description, lip-sync yes/no, and character or location references. The general format is in Storyboards, Animatics, Shot Lists and Overheads.
Then check each row against what the model can generate in one pass:
Model
Max clip per generation
4-bar phrases per clip at 120 BPM (our arithmetic)
Source
Veo 3.1
8 s
1
Google model page
Kling 4.0
30 s
3.75
Kling 4.0 user manual
Kling 4.0 Flash
20 s (720p)
2.5
Kling 4.0 user manual
Longer limits don't mean longer shots. Kling 4.0 has smart shot-splitting on by default, per its user manual, so a 30-second generation may come back as cuts the model chose, not cut to your beat. For tightly cut performance shots, generate a little longer than the slot and trim.
Step 4: How do you generate the shots?
Start from images for the performer. A music video lives on a recognisable artist across dozens of shots, and vendors recommend image-to-video or reference images for that. Runway positions text-to-video for when "exact character or scene consistency isn't the priority" (see Text to Video vs Image to Video).
Generate silent footage where the model allows it. You'll replace the audio anyway, and silent is cheaper:
Kling 4.0 keeps audio-visual sync on with "no toggle to turn them off"; write "no background music" so the audio is easier to strip.
Prompt for rhythm in the picture. Neither model can hear your track, but Google's Veo guide supports timestamp prompting ([00:00-00:02] …) and Kling's manual recommends "X–Y s" segments. Matching segments to beat times gives the edit natural cut points; syntax is in the Veo 3 Prompt Guide and Kling Prompt Guide.
Step 5: What are the lip-sync options for an AI music video?
Lip-sync tools take your audio and either rework the mouth in an existing video or animate a still. Published options as of 18 November 2026:
Tool
Input
Audio limits
Published price
Source
Kling Lip Sync
An existing video with a detected face, plus audio
2–60 s per audio segment; mp3/wav/m4a/aac up to 5 MB; one person per task
No documentation we read promises anything specific about singing (held notes, vibrato, wide-open vowels), so test one chorus before committing a budget. Runway describes Act-Two only as its "motion capture model" without saying whether it carries mouth movement; test that too. Our suggestion: feed the tool an isolated vocal, not the full mix, so drums don't drive the mouth. Suno's paid plans include stem separation, and Runway's API lists an ElevenLabs voice-isolation model at about $0.0017 per second.
Step 6: How do you edit an AI music video to the beat?
Lock the song on the timeline, add the beat markers from Step 2, then cut picture to markers.
Cut on structure, not just beats. Oliveira's four-bar colour-coding gives you phrase starts. We suggest saving changes of location or look for section changes, such as verse to chorus.
Vary cut density by section. Cutting every beat at 140 BPM means a new shot every 10 frames. Hold longer in verses and save the fastest cutting for peaks; see how editors decide where to cut.
Keep lip-sync shots in sync. Never slip a lip-synced clip off its position against the song; slide the cut points around it.
Match the look across models. Grade shots from different models to a shared reference (Grading to a Reference Frame).
Any editor with markers will do; our AI video editors comparison lists prices and export limits. Vinc's landing page describes "a real timeline with video, voice, effects and music lanes, a mixer, captions and export", with image, video, voice and music models "picked per shot".
What does an AI music video cost?
Our arithmetic from list prices checked 18 November 2026: a 3 minute 30 second song (210 s), twice as much footage generated as used (420 s), and 60 seconds of lip-sync.
Generation cost for 420 seconds of 1080p footage
Kling 4.0$99
Veo 3.1 (silent)$84
Kling 3.0 (silent)$47
Veo 3.1 Fast (silent)$42
Our arithmetic: 420 s × list price per second, checked 18 Nov 2026. Silent output where the model offers it (Veo 3.1 $0.20/s, Veo 3.1 Fast $0.10/s, Kling 3.0 $0.112/s); Kling 4.0 includes audio ($0.235/s). Single price per attempt; excludes music, lip-sync and editing.
Add lip-sync: $0.84 on Kling Lip Sync (12 × $0.07), or $3.00 on sync.so lipsync-2 plus a $19 Creator month (Hobbyist watermarks output). Add one month of Suno Pro at $10. A 2:1 footage ratio is optimistic, and retries scale only the generation line; see What an AI Shot Really Costs.
FAQ
Which aspect ratio should a music video use?
Use 16:9 for YouTube and 9:16 for Shorts, Reels and TikTok. Veo 3.1 outputs both, and Kling 4.0 adds 21:9, 1:1, 4:3 and 3:4. See Aspect Ratios Explained.
How we compiled this
On 18 November 2026 we read the vendor documentation and price lists below; editing practice comes from No Film School's interview with Leonardo Oliveira. The beat table, cost chart and worked example are our arithmetic, and we did not test any lip-sync tool on singing. We left out Hedra because its pricing page did not say what its credits buy in seconds of video; Vinc makes an AI film studio.
AI Music for Film: Suno, Udio and Other Generators on Licences, Ownership, Lawsuits and Cue Sheets
Whether an AI-generated track can go in your film depends on the plan you made it on, whether the terms exclude film and TV, and whether anyone can own it at all. This guide compares the published terms of Suno, Udio, ElevenLabs Music, Stable Audio and AIVA, sets out the label lawsuits as reported, and explains what US copyright law and North American PRO rules mean for your cue sheet.