I Made Motion Graphics Using Only Claude Code & Hyperframes
A full breakdown of an AI video-editing pipeline where Claude runs Premiere Pro directly, cutting footage and building motion graphics from a chat prompt.
Posted
1 months ago
Duration
Format
Tutorial
educational
Views
6.4K
137 likes
57 · 43
Big Idea
The argument in one line.
AI can now execute the technical side of video editing end to end, from rough cuts to custom motion graphics, but it only replaces execution, not the taste and creative direction that decide whether the output looks good or generic.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You edit your own videos and want to cut the hours spent scrubbing timelines and fighting After Effects keyframes.
You're comfortable installing open-source tools and following GitHub setup instructions, or want a pre-built folder you can just plug in.
You already have a sense of pacing, storytelling and visual hierarchy and want a tool to execute that vision faster.
You use Premiere Pro, DaVinci Resolve Studio, or Final Cut Pro and want an AI agent that can control the app directly instead of just generating clips.
SKIP IF…
You're looking for a one-click consumer app — this requires setting up local tools, API access, and an MCP bridge into your editor.
You have no editing taste or direction yet — the video is explicit that vague prompts like "make it cool" produce generic output.
TL;DR
The full version, fast.
A creator built a custom AI editing system where Claude Code, running autonomously inside Premiere Pro through an MCP bridge, does the technical work of editing. WhisperX transcribes footage with word-level timestamps so Claude knows where to cut. Hyperframes lets Claude generate motion graphics by writing HTML/CSS, screen-recording the animation, and dropping the result onto the timeline. In a live demo, a single prompt turns 10 minutes of raw footage into a 4-minute rough cut, and a full animated graphic gets built, refined by follow-up prompts, and placed on the timeline in about 30 minutes versus 3+ hours by hand. The takeaway: AI removes the execution bottleneck, but the creative direction, i.e. what to ask for and how to refine it, still has to come from the editor.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
States the premise: every cut and graphic in the video was made by Claude running autonomously inside Premiere Pro, and frames the technical friction of editing as creativity's biggest bottleneck.
01:09 – 02:51
02 · How it works: WhisperX, Hyperframes & the Premiere Pro MCP
Breaks down the three-agent stack: WhisperX transcribes with word-level timestamps, Hyperframes generates motion graphics by writing and screen-recording HTML/CSS animations, and the Premiere Pro MCP lets Claude place clips and apply effects directly on the timeline. Notes MCPs also exist for DaVinci Resolve Studio and Final Cut Pro.
02:51 – 03:47
03 · Live demo: the rough cut from one prompt
Types a single prompt asking for a hard rough cut of a 10-minute raw clip with standard audio presets, then steps away while Claude condenses it into a paced 4-minute cut with leveled audio.
03:47 – 04:41
04 · Live demo: AI motion graphics + refining the look
Describes a graphic (background, typography, frame rate) and Claude builds it in about 16 minutes, keying PNGs and adding drop shadows. Refines it with follow-up prompts (move text, change accent color, add a bouncing mascot) in about 3 more minutes.
04:41 – 05:28
05 · Music, export & the real time savings
Adds a background track and sets its level by voice-safe decibel target, exports to a separate output folder without overwriting the source project, then states the full workflow took about 30 minutes versus an estimated 3+ hours by hand.
05:28 – 06:36
06 · How to get it (DIY setup or pre-built download)
Offers two paths: paste the video's own transcript into Claude as a build guide for a few hours of DIY setup, or download the creator's pre-built folder (Hyperframes code, MCPs, audio presets, custom instructions) for a plug-and-play system.
06:36 – 08:26
07 · Will AI replace video editors? Taste over execution
Argues AI replaces execution, not creativity: vague prompts produce generic output, and the editors and engineers getting the best results are the ones who already have taste and use AI to type, not to think.
Atomic Insights
Lines worth screenshotting.
The system's three-agent stack is WhisperX for word-level transcription, Hyperframes for AI-generated motion graphics, and an MCP bridge that lets Claude control the editing app's timeline directly.
Claude generates motion graphics by writing HTML and CSS to build an animated webpage, then the system screen-records that page and drops the resulting video file onto the timeline.
A single prompt turned a 10-minute raw clip into a paced 4-minute rough cut with leveled audio, with no manual timeline scrubbing.
A full mixed-media animated graphic, keyed PNGs, drop shadows and all, was built and placed on the timeline from one descriptive prompt in about 16 minutes.
Refining an existing graphic (moving text, changing accent color, adding a bouncing mascot image) took about three additional minutes of follow-up prompting.
The entire live-demo workflow, cut plus graphic plus music plus export, took about 30 minutes versus an estimated 3+ hours to do the same work by hand in After Effects.
MCP bridges exist for Premiere Pro, DaVinci Resolve Studio, and Final Cut Pro, so the same control pattern isn't locked to one editor.
The creator argues AI replaces execution, not creativity: a vague prompt like "edit my video and make it cool" produces generic output, while a specific, taste-driven prompt produces a result worth shipping.
The comparison drawn is to software engineers who use AI to type rather than to think — the tool multiplies existing taste, it doesn't substitute for having none.
Takeaway
AI removes editing's technical friction, not its creative direction
WHAT TO LEARN
A three-tool pipeline can now transcribe footage, build motion graphics from a text description, and place everything on a timeline automatically, but the output is only as good as the specific creative direction fed into it.
Word-level transcription is what lets an agent know exactly where dead air and mistakes are, which is the foundation any automated cutting decision depends on.
Generating a video asset doesn't require a video-generation model: writing an animated HTML/CSS page and screen-recording it is a viable, controllable way to produce a motion graphic.
Direct app control (an agent placing clips and applying effects on a real timeline) is a different capability than generating a rendered clip elsewhere and importing it, and it's what actually removes the manual editing step.
A specific, detailed prompt (background, typography, frame rate) produced a usable result; a vague prompt like "make it cool" is explicitly called out as producing generic, boring output.
Refining a generated asset with follow-up instructions (reposition an element, change a color, add and animate a new image) is fast, on the order of a few minutes, which makes iteration cheap once the first draft exists.
The stated time comparison, about 30 minutes end to end versus 3+ hours by hand, assumes the editor already knows what they want; the time saved is technical execution time, not decision-making time.
The same control pattern (an agent driving the app via a protocol bridge) generalizes across editors, since MCPs are noted for DaVinci Resolve Studio and Final Cut Pro as well as Premiere Pro.
Glossary
Terms worth knowing.
Hyperframes
A tool that lets an AI agent generate motion graphics by writing HTML and CSS for an animated webpage, then screen-recording that page into a video file.
MCP (Model Context Protocol)
A bridge that lets an AI agent like Claude directly control a desktop application, such as placing clips and applying effects inside Premiere Pro's timeline.
WhisperX
An open-source transcription tool that generates word-level timestamps from audio, used here to tell the AI exactly where dead air and mistakes are so it knows where to cut.
Rough cut
A first-pass edit that trims raw footage down to its essential parts before finer editing, color, and sound work happens.
“The technical friction of editing is the biggest bottleneck to your creativity.”
Sharp reframe of a common creator complaint→ TikTok hook↗ Tweet quote
06:43
“This AI video editor doesn't replace creativity. It replaces execution.”
Clean thesis statement, works as a standalone claim→ IG reel cold open↗ Tweet quote
07:11
“If you just open cloud and type edit my video and make it cool, you are going to get a very boring generic output.”
Concrete warning against lazy prompting→ newsletter pull-quote↗ Tweet quote
07:32
“The engineers building the best AI apps aren't letting AI do the thinking. They use AI to do the typing.”
Transfers a software-engineering framing directly onto video editing→ TikTok hook↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphoranalogy
If you make videos, you already know the hardest and most annoying part of the process isn't coming up with ideas. It is the execution. It is sitting in a dark room for even hours or even tens of hours, scrubbing through the timelines, cutting out dead hair and fighting with After Effects keyframes just to make a simple graphic pop.
The technical friction of editing is the biggest bottleneck to your creativity. And you're not alone. I struggle with this too.
But what if the technical side of editing was completely stripped out? What if you and I could just describe what we want to see and AI builds it in the timeline? Well, every single cut, transition and motion graphic you're watching on the screen right now was not edited by me.
It was entirely edited by Claude. running autonomously inside Premiere Pro. I've spent the last few weeks custom building an AI video editing workflow that does exactly that.
And today I'm taking you under the hood. I'm going to show you exactly how it works, how it transforms you from an editor to a director, and most importantly, how you set it up for yourself. So let's dive in.
Let's break down how this actually works. The entire brain of the system is just a cloud project folder living on my computer. It relies on three core agents to work.
The first is WhisperX. Before Cloud can edit a video, it needs to understand what is actually being said and when. Whisper is an incredibly powerful open -source transcription tool.
We use it to transcribe the raw video and generate word -level timestamps. This is the years of the operation. It tells Cloud exactly where the data error is, where the mistakes are, and where the cuts need to happen.
The second is called hyperframes. This is how the AI actually generates motion graphics. Now, you might be wondering, how is Cloud generating video files?
It's actually a brilliant workaround. We already know Cloud is incredible at writing code and building websites. So, when I ask for a graphic, Cloud writes HTML and CSS to build a fully animated webpage.
It makes the text slide, the image pops, and the color shifts. then the system simply screen records that web page animation, turns it into a video, and drops it into your folder, Premiere Pro MCP. If you haven't heard of MCPs yet, it's essentially a bridge that lets Cloud directly control your apps.
So Cloud makes the graphic, uses the MCP to reach into Premiere Pro, places the clip on your timeline, makes the cuts, and applies the effects. And for those of you who are not on Premiere, the video Ninja Resolve Studio and Final Cut Pro also have their own MCPs.
I want you to see exactly how this feels to use. Let's do a live demo, starting with the most seditious part of any videos – the raw cut. Normally, you would sit here scrubbing through 10 minutes of raw footage.
Instead, I'm going to grab the file path of my raw video, drop it into cloud, and just talk to it like it's my assistant. I'll type I need a rough cut for this footage, cut it hard, only keep the absolute essential parts, drop it in the Premiere timeline right after the intro and apply my standard audio presets.
I hit enter and now I can step away. go make a coffee or start writing my next video. Alright, so it's taken a 10 -minute clip, condensed it into a perfectly paced 4 -minute rough cut and leveled the audio.
But where this really shines is the graphics. Let's say I want to recreate that animation you saw earlier. I just go to cloud and describe my vision, the format like background, typography and technical aspects like frame rate.
16 minutes later it's done. It actually went out, found the PNGs, keyed them out, built the animation, added the drop shadows and placed it in my timeline. Now, this next step is crucial because this will make your videos look actually good instead of generic AI slop.
So I review the graphics and refine it by typing, move the text to the bottom, change the accent color to cloud orange, add an image of Claude Mascot and make it bounce and glow. About three minutes later, it updates the graphic entirely.
To finish it off, I just paste a link to a track on my hard drive and say, add this as a background music and set audio level to minus 23 decibels so my voice doesn't get faded. Finally, I ask it to export and it outputs the final file directly to my downloads folder without overwriting my original project file. I want to be completely transparent with you on the time savings here.
All in, with the back and forth, prompting and iterating, this took me about 30 minutes. To edit this video normally, with this exact graphics, it would have taken me at least 3 hours. And that is assuming I even know how to build those animations in After Effects from scratch, which would take days to learn.
Now I want you to have access to this. There are two ways to get it running on your machine. The first way is for those who want to set everything up themselves and actually understand what is happening under the hood.
All you have to do is copy the transcript of this YouTube video, paste it into a BlendCloud project and tell Cloud, use this transcript as a guide to help me build the entire video editing system. It will walk you through the GitHub repos, the installations and the code. It will probably take you a few hours of tinkering and troubleshooting to get it dialed in, but it's 100 % doable.
The second way is for those of you who want to skip the setup and just get straight to creating. I've taken the exact folder I use, with all the hyperframed codes, the MCPs, the audio presets and the custom instructions already wired up, and I put it in the link below. You can just download the folder, plug it in, And it works right out of the box.
I'll leave the link to that right at the top of the description. I want to zoom out for a second, because whenever I show this to people, I inevitably get this question. Is AI going to replace video editors?
I want to reframe how we think about this. The short answer is no. But it is going to fundamentally change what your job actually is.
This AI video editor doesn't replace creativity. It replaces execution. execution.
The only reason that graphic looked so good is that I had the specific idea to ask. If you just open cloud and type edit my video and make it cool, you are going to get a very boring generic output. It will just output what it considers cool.
When people get frustrated with AI and say, it didn't do a good job, it is actually because they are trying to outsource their ideas to the AI. And honestly, this is the worst way to use this technology. Think about software engineers right now.
The engineers building the best AI apps aren't letting AI do the thinking. They use AI to do the typing. They already have a deep understanding of what makes an app great.
They have a taste. It is the exact same. for video editing.
If you already have a great taste, if you understand pacing, storytelling, and visual hierarchy, this tool is going to give you superpowers. You no longer have to limit your ideas to your technical ability and after -effects. AI is not here to replace you.
It is here to give you your time back, so you can actually spend it on the ideas that matter. Don't overhaul your entire workflow today, just pick one thing and try automating it. it.
If you want to bypass this setup, just download my pre -built system to do exactly that, check out the link in the description. So, until next time guys, keep creating!
The Hook
The bait, then the rug-pull.
The pitch is blunt: every cut, transition and motion graphic in the video wasn't made by the creator, it was made by Claude running autonomously inside Premiere Pro. What follows is the under-the-hood breakdown of how that system works and a live demo of it cutting a rough cut and building a graphic from a chat prompt.
Frameworks
Named ideas worth stealing.
01:16list
The three-agent editing stack
WhisperX (transcription + word-level timestamps)
Hyperframes (HTML/CSS-generated motion graphics)
App MCP (direct timeline control in Premiere Pro / Resolve / Final Cut)
The three tools that together let Claude go from raw footage to a finished, graphics-included edit without the editor touching the timeline.
Steal forAny local Claude Code automation pipeline that needs an agent to both understand media content and directly manipulate a third-party desktop app.
CTA Breakdown
How they asked for the click.
VERBAL ASK
06:35link
“I've taken the exact folder I use, with all the hyperframed codes, the MCPs, the audio presets and the custom instructions already wired up, and I put it in the link below.”
Soft-pitched as a time-saving shortcut rather than a hard sell, offered as the second of two explicit options (DIY vs. pre-built download) so it doesn't feel forced.
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A GitHub 'skill' hands Claude Opus 5.5 a talking-head clip and its transcript, and it plans, times, and renders word-synced B-roll animations without being told what to draw.
A five-minute demo where an AI coding agent copies any video's edit style off a single YouTube link and rebuilds it in code, no video-generation credits required.
A creator wires Claude Code into DaVinci Resolve through a free MCP bridge and runs every stage of an edit live on camera, rough cut, motion graphics, and color grade, with a timer running so you can see exactly how long each pass actually takes.
A screen-recorded, un-sped-up walkthrough of five real motion-graphics jobs — rough cut, animated background, number counter, typography titles, Vox-style collage — built entirely by prompting Claude Code and letting it drive CapCut.
A YouTuber hands Claude full control of Premiere Pro — cutting a rough cut and generating an HTML-animated Mars/Earth graphic from prompts alone — then walks through building the same system for free.
A creator wires Claude Code into DaVinci Resolve through an open-source MCP server, then stress-tests whether it can actually follow marker-based, relative instructions the way a real junior editor would.