How to make an AI music video: from track to finished edit
No crew, no location, no hire van. An AI music video swaps the shoot for prompts, renders and a proper edit. Here is the whole job, from the locked track to the post.
Published 2026-10-26 by Cat Claw, the UK's create-and-distribute AI studio. About a 16 minute read.
The short version
- An AI music video replaces filming with generated shots: you prompt each scene, render short clips and edit them to your track in the usual way.
- AI video models render 1 clip at a time, from 3 to 30 seconds depending on the model, so a 3-minute video is built from dozens of shots, not from a single prompt.
- Consistency comes from a style sheet and approved reference images reused in every shot. Reference-to-video models such as Seedance 2.5 and MiniMax H3 accept images, video and audio as references.
- On Cat Claw AI, a 10-second MiniMax H3 clip costs 50 credits at the standard rate, a song from Music Studio costs 8 credits, and a failed render refunds automatically.
- Export a 16:9 master for YouTube and a 9:16 cut for TikTok, Instagram Reels and YouTube Shorts. YouTube recommends MP4 files with H.264 video.
- To join the YouTube Partner Programme, a channel needs 1,000 subscribers plus either 4,000 public watch hours in 12 months or 10 million Shorts views in 90 days.
- UK law currently protects wholly computer-generated works under the CDPA 1988, but the government proposed removing that protection in March 2026, so keep a record of your own creative input.
An AI music video is a music video whose visuals are generated by AI image and video models instead of filmed with a camera, crew and location. Each shot starts as a written prompt, a reference image or both, renders as a short clip, and is then cut to the track in an editor like any other footage.
This guide covers the whole job: what you need, how to plan shots against a track, how to keep a character consistent, how to sync and edit, what it costs in credits, how to export and post it, and how it can earn. It is written by Cat Claw AI, a UK-based AI creative studio that has produced 2 full music videos for the artist Skylark, and every product detail in it has been checked against the studio as it works today.
What is an AI music video?
Quick answer. An AI music video is a music video made from generated visuals rather than filmed footage. Each shot is described in a prompt, often anchored by reference images, and an AI video model renders it as a clip. The clips are edited to the song like any other footage. The track can be AI-made or recorded by hand; the AI part is the pictures.
What changes is where the effort goes. A shoot spends its budget on crew, kit, locations and days on set. An AI production spends it on the concept, the prompts, the retakes and the edit. The craft doesn't disappear. It moves from the set to the screen.
How is an AI music video different from a traditional shoot?
| Category | AI music video | Traditional shoot |
|---|---|---|
| Crew and location | None needed (leads) | Director, crew, cast, locations and kit |
| What drives the cost | Credits per rendered second, plus your time | Day rates, hire and post-production |
| Changing a shot | Rewrite the prompt and render again (leads) | Reshoot, or fix it in post |
| Visual range | Any setting, era or style in 1 project (leads) | Limited by budget and physics |
| Real performance | Generated, and lip sync needs checking shot by shot | Captured on camera (leads) |
| Same face in every shot | Needs references and discipline | Same actor in every take (leads) |
Which songs suit an AI music video?
Almost any genre works, but some briefs play to the medium's strengths: surreal or dreamlike concepts, period or sci-fi worlds a shoot could never afford, stylised animation, and narrative videos told through a recurring character.
Word-perfect singing across a whole song is still the hardest thing to generate. Plan a story with cutaways rather than 3 minutes of close-up lip sync, and the medium works with you instead of against you.
What do you need to make an AI music video?
Quick answer. You need a finished track (or a demo with its final structure), an image generator for keyframes and references, an AI video generator, an editor that can hold the full song, and somewhere to post it. You don't need a camera, a crew or a powerful computer, because the rendering happens in the cloud.
| Job | What it does | In Cat Claw AI |
|---|---|---|
| The track | The song the video is cut to, in its final arrangement | Your own recording, or a song from Music Studio: MiniMax Music makes full songs up to 5 minutes, and Cat Squawk sits alongside it, at 8 credits a song |
| Keyframes and references | Stills that fix a character, set or palette before anything moves | Image models such as Seedream, with CatClaw Layers for editing |
| Video shots | Turns a prompt, a still or a set of references into a clip | Video models including Seedance 2.5, MiniMax H3, Kling 4.0 and Wan 3.0 |
| The edit | Puts clips in order under the song, trims them and adds titles | Cut Studio for teasers and social cuts; a desktop editor for the full-length master |
| Posting | Publishes the finished cuts to your channels | Puurfect Post, to around 15 social platforms, at no credit cost |
Do you need technical skills to make an AI music video?
No coding, no 3D and no camera work. You need to describe a shot clearly, judge whether a render works, and run a basic edit: order clips, trim them and line cuts up with the music.
If you've ever cut a video on your phone, you have enough to start. The prompts and the edit both get sharper with every project.
Which AI video models suit a music video?
The models that suit music videos are the ones that take references, because references are how you keep a world consistent. These 3 cover most of the job on Cat Claw:
| Model | Clip length | Output | Best use in a music video | 10-second clip |
|---|---|---|---|---|
| Seedance 2.5 (reference to video) | 4 to 30 seconds | 480p or 720p | Up to 30 images, 10 videos and 10 audio clips as references; audio-driven shots such as lip sync | 68 credits |
| MiniMax H3 | 4 to 15 seconds | 768p or 2K, stereo sound | Cinematic and performance shots; the model behind the Skylark videos | 50 credits |
| Kling 4.0 (keyframes) | 3 to 30 seconds | 720p or 1080p | Controlled moves between your own first, middle and last frames | 56 credits |
Prices are the standard rate before any sale, and every render shows its credit price before you confirm. For the wider field, model by model, see the comparison below.
How to make an AI music video, step by step
Making an AI music video follows the same order as a shoot: concept, plan, shoot, edit, deliver. The difference is that every shot is a render you can repeat. Work in this order and your credits go on shots you'll actually use.
- Lock the track. Finish the song, or at least its structure, before you render anything. Note the timestamp of every section: intro, verses, choruses, bridge and drop. Every later decision hangs off this map.
- Write a 1-line concept. Say what the video is about in a sentence, and whether it's narrative, performance, abstract or a mix. "A singer chasing a signal through a rain-soaked city at night" is a concept. "Cool visuals" isn't.
- Build a style sheet. Fix the palette, lighting, lens, era, wardrobe and setting in a short block of text you'll paste into every prompt. Add 3 to 6 reference stills that show the look.
- Design your character. If someone recurs, generate a reference set first: face, full body and wardrobe, from more than 1 angle. Those images go into every shot the character appears in.
- Write a shot list against the timestamps. List each shot with its section, length, framing and action. Plan more shots than the song needs, so the edit has choices.
- Render keyframes, then animate. Make a still for each key shot and fix it until it's right, then animate it with an image-to-video or reference-to-video model. A still costs from 3 credits; a clip costs far more, so settle the look on stills.
- Edit to the music. Lay the full track on the timeline, drop clips onto their sections and cut on the beat. Mute the clips' own generated sound unless you want it in the mix.
- Export and post. Export a 16:9 master and a 9:16 cut, watch both on a phone, then post. Write titles, descriptions and captions before release day, not on it.
How do you keep a consistent visual style across an AI music video?
Through repetition, not luck. Paste the same style block into every prompt, keep the camera language consistent, and start shots from your own approved keyframes rather than a blank prompt.
Reference-to-video models make this much easier. Seedance 2.5 takes up to 30 reference images in 1 render, and MiniMax H3 mixes image, video and audio references in a single request. Change 1 thing at a time between retakes, or you won't know which word fixed the shot.
How many shots does a 3-minute music video need?
More than most people plan for. If you cut every 3 to 4 seconds, a common pace for a music video, a 3-minute track needs 45 to 60 shots.
Render clips longer than you'll use and trim them in the edit. The strongest 3 seconds of a clip are rarely its first 3.
How do you sync AI visuals to the music?
Quick answer. Sync happens in the edit, not in the prompt. Lay the full song on the timeline, mark each section, place clips on their section, then move each cut onto a beat. Save the biggest visual changes for section changes, such as the first chorus or the drop. Fast cuts suit high-energy sections; longer shots suit quiet ones.
Cutting on the beat
Put cuts on the kick or the snare, not between them. Most editors draw the audio waveform, which is usually enough to see where the beats land.
A cut that lands a frame late feels wrong even when nobody can say why. Zoom in on the timeline for the big moments and get them exact.
Performance and lip sync shots
Generated singing is the hardest shot to get right, so use it where it counts. Seedance 2.5's reference-to-video mode accepts up to 30 seconds of reference audio per render, so you can give it the exact slice of the track a shot covers.
Check every lip sync take at full size before it goes into the edit. Cover weak moments with cutaways rather than forcing a shot that nearly works.
Can you edit a full AI music video in Cat Claw?
Partly, and it's worth being precise. Cut Studio, the free editor inside Cat Claw, holds up to 30 clips per cut at 9:16, 16:9 or 1:1, with trims, captions, title cards, beat-synced cuts and a hook check. Its music comes from a library of 4 beds, so today it can't lay your own full-length song under an edit.
That makes Cut Studio a good fit for teasers and social cutdowns made from your clips, and a desktop editor the right tool for the full-length master with your own track. Clips download from your gallery clean, with no watermark, so moving between the 2 is simple.
How much does it cost to make an AI music video?
Quick answer. On Cat Claw AI, 60 ten-second MiniMax H3 clips cost 3,000 credits at the standard rate, which is the 3,000-credit pack at £109.99. That is enough shots for a 3-minute video with retakes. Keyframe stills start at 3 credits, a song from Music Studio is 8 credits, and editing in Cut Studio and posting cost nothing.
| Item | Credits | Notes |
|---|---|---|
| 1 song in Music Studio | 8 | Per song, on either engine |
| 1 keyframe still | From 3 | Image renders start at 3 credits |
| 10-second MiniMax H3 clip | 50 | 5 credits a second |
| 10-second Kling 4.0 clip | 56 | Keyframe control |
| 10-second Seedance 2.5 clip | 68 | Image, video and audio references |
| 60 ten-second MiniMax H3 clips | 3,000 | A 3-minute video with room for retakes |
| Editing in Cut Studio | 0 | Optional AI fixes on a clip are the only paid extras |
| Posting with Puurfect Post | 0 | Around 15 social platforms |
Credit packs run from 100 credits for £4.99 to 3,000 credits for £109.99, and pack credits last 12 months. Monthly plans run from Starter at £14.99 (350 credits) to Agency at £199.99 (5,400 credits), and plan credits reset each month.
A failed render refunds automatically, so you pay for clips that come back, not for errors. You still pay for clips you reject, which is why the keyframe step matters: settle the look on 3-credit stills, then animate only what you've approved.
How do you export an AI music video for YouTube, TikTok and Instagram?
Quick answer. Export a 16:9 master for YouTube and a 9:16 vertical cut for TikTok, Instagram Reels and YouTube Shorts. YouTube recommends an MP4 file with H.264 video and also accepts vertical and square uploads. Plan the vertical cut from the start, because cropping a wide shot later loses the edges of the frame.
| Platform | Aspect ratio | What to export | Use it for |
|---|---|---|---|
| YouTube | 16:9 | MP4 with H.264 video, at the resolution you rendered | The full-length master and the premiere |
| YouTube Shorts | 9:16 | MP4 | The hook or chorus, cut vertical |
| TikTok | 9:16 | MP4 | Short vertical cuts that point back to the full video |
| Instagram Reels | 9:16 | MP4 | The same vertical cuts, posted alongside TikTok |
Render your short-form shots vertical
Seedance 2.5 and MiniMax H3 both render 9:16 natively as well as 16:9. Rendering your key short-form shots in 9:16 gives a cleaner vertical cut than cropping the wide master.
For the rest, Cut Studio renders a 9:16 edit and fits wide clips with blur, crop or letterbox, so a single set of clips can still feed both formats.
The first 3 seconds of a vertical cut
On short-form feeds, the opening decides whether anyone watches the rest. Open on your strongest image or the hook of the song, not the intro.
Cut Studio's hook check scores the first 3 seconds of a cut and suggests fixes in plain English before you post.
Posting to every platform at once
Puurfect Post publishes a finished file from your Cat Claw gallery to around 15 social platforms in 1 tap. Posting costs no credits, a post can be scheduled for release time, and nothing goes out without your approval.
Can you make money from an AI music video?
Quick answer. Yes. An AI music video can earn YouTube ad revenue once the channel joins the YouTube Partner Programme, drive streams of the track, win commissioned and brand work, and feed a catalogue of prompts you sell. The video rarely pays on its own on day 1. It's the asset that sends people to everything else.
- 1,000 subscribers needed to apply to the YouTube Partner Programme (YouTube Help Centre, October 2026)
- 4,000 public long-form watch hours in the past 12 months, the long-form route (YouTube Help Centre, October 2026)
- 10 million Shorts views in the past 90 days, the Shorts route (YouTube Help Centre, October 2026)
Ways an AI music video earns
- YouTube ad revenue, once the channel meets 1 of the Partner Programme thresholds and passes YouTube's review.
- Streams of the track, with the video and its vertical cuts sending listeners to your release link.
- Prompt sales: every prompt on the Cat Claw marketplace sells for £1.99, and the seller earns 20 credits a sale.
- Licensing the music itself, through royalty-free and sync licences.
- Commissions: a finished video is the strongest portfolio piece for artists and brands who want one made.
- How to create and sell AI-generated music online
- How to sell AI prompts: pricing, listings and what actually sells
Who owns an AI music video in the UK?
Under the Cat Claw Terms you own your outputs, to the extent ownership can exist in AI-generated material, and you can use them commercially. The wider law is less settled.
The Copyright, Designs and Patents Act 1988 protects wholly computer-generated works and treats the person who made the arrangements for the work as its author. The government's Report on Copyright and Artificial Intelligence (18 March 2026) proposed removing that protection, while keeping copyright for works made with AI assistance. In the US, the Copyright Office has said material generated purely by AI, without human creative input, can't be registered.
Either way, your human contribution is what counts, so keep your concept, shot list, prompts, keyframes and edit decisions. The song carries its own rights too, so make sure you hold them before you publish. This is general information, not legal advice.
What does a real AI music video production look like?
Cat Claw has produced 2 full music videos for Skylark, released by Winwick Records and VoxaRox in association with Cat Claw AI. Static, the debut featuring Dexter Koh, and Voicemail, its sequel, were both made through an AI filmmaking workflow in Cat Claw, with MiniMax H3 generating the performance and cinematic sequences and no traditional set or crew.
The pipeline ran in the same order as the steps above. The human team set the concept and story, then the character, then a visual identity of wardrobe, palette, lighting and lens carried through every shot. Performance and cinematic sequences followed, then the edit, then cutdowns for social.
The part worth copying is the sequel. Voicemail brings back the same character and the same world as Static, which is the consistency problem solved twice, across 2 separate productions. Human creative direction made every significant call; the models rendered and people decided.
- Static by Skylark: inside its AI-generated music video
- Voicemail by Skylark: the sequel single and its AI music video
If your first renders look like something the cat dragged in, that's normal. Change 1 thing, render again, and keep the shot list moving.
Frequently asked questions
What is an AI music video and how is it different from a normal one?
An AI music video uses generated visuals instead of filmed footage. Each shot is written as a prompt, often with reference images, and an AI video model renders it as a short clip, which is then edited to the track. A normal music video is shot on camera with a crew, cast and locations. The edit, the concept and the creative judgement are the same in both.
Can I make an AI music video for free?
You can start for free, but the video shots themselves cost credits. Sign-up to Cat Claw AI is free with no card, editing in Cut Studio costs no credits, and posting through Puurfect Post costs nothing. Rendering is the paid part: a 10-second MiniMax H3 clip is 50 credits at the standard rate, and the smallest credit pack is 100 credits for £4.99.
How long does it take to make an AI music video?
It depends far more on planning and editing than on rendering. Each clip on the models above renders in a few minutes, so the machine time for a 3-minute video of 45 to 60 shots is measured in hours. Concept, retakes and the edit take most of the calendar. A simple performance video comes together faster than a narrative piece with a recurring character.
Do I need editing experience to make an AI music video?
No, but you do need to edit. The core skills are ordering clips, trimming them and placing cuts on the beat, and any phone or desktop editor that shows the audio waveform will do it. Cut Studio inside Cat Claw handles teasers and social cuts for free. For the full-length master with your own song, a basic desktop editor is enough.
What is the best AI tool for making music videos in 2026?
The best tool is the one that takes references, renders long enough clips and fits your edit. For music videos that usually means a reference-to-video model such as Seedance 2.5 or MiniMax H3. Cat Claw AI puts both, plus Kling 4.0, Music Studio for the track and Puurfect Post for distribution, on 1 credit balance, with a desktop editor for the full-length cut.
Can I use AI-generated visuals commercially in my music video?
Usually yes, but it depends on the tool's terms. Under the Cat Claw Terms you own your outputs, to the extent ownership can exist in AI-generated material, and you can use them commercially. Copyright in AI-generated work is unsettled in the UK and the US, so keep records of your creative input and take legal advice for big commercial deals. This is not legal advice.
How do I keep the same character throughout an AI music video?
Design the character first as a set of reference images, covering face, full body and wardrobe from more than 1 angle, then feed those references into every shot the character appears in. Reference-to-video models such as Seedance 2.5 and MiniMax H3 are built for this. Keep the same style block in every prompt and change 1 thing at a time between retakes.
Can I post an AI music video on YouTube and TikTok at the same time?
Yes. Export a 16:9 master for YouTube and a 9:16 vertical cut for TikTok, Instagram Reels and YouTube Shorts, then post them together. On Cat Claw, Puurfect Post publishes a finished file from your gallery to around 15 social platforms in 1 tap, can schedule the release time, and costs no credits. Nothing posts without your approval.
Make your first AI music video.
The track, the keyframes, the video shots and the posting, on 1 credit balance in 1 UK studio. Sign-up is free with no card, and failed renders refund automatically.
How to make an AI music video from scratch: plan shots to your track, keep a character consistent, cut to the beat, post everywhere and see what it costs.