06:14Jamie Fenn · Tutorial
DaVinci Resolve just made it 10x FASTER
A DaVinci Resolve editor hands Claude the whole editing brief and gets back motion graphics that used to take Fusion hours to build by hand.
September 25thThe argument in one line.
HyperFrames turns plain-English prompts into real rendered motion graphics inside DaVinci Resolve, but the tool was never the hard part, taste is, and that taste is seven repeatable rules anyone can paste into a prompt.
Read if. Skip if.
READ IF YOU ARE…
- You edit video and want motion graphics without learning After Effects or Fusion nodes.
- You've tried AI graphic generators and gotten generic-looking results you didn't want to use.
- You use Claude Code or ChatGPT already and want to extend it into your video editing workflow.
- You edit in DaVinci Resolve and want graphics that land directly on your timeline.
SKIP IF…
- You already own a full motion-graphics pipeline in After Effects and have no interest in prompt-driven tools.
- You don't edit in DaVinci Resolve and aren't willing to export clips to a separate chat tool.
The full version, fast.
HyperFrames is a free, open-source tool that lets Claude Code or ChatGPT render real motion-graphic video files from plain-English prompts straight into a DaVinci Resolve timeline. Installed in four steps, it works instantly, but the default output looks like every other AI graphic: generic colors, uniform sizing, nothing guiding the eye. The fix isn't a better tool, it's seven rules pasted into the prompt: one locked visual look, every graphic timed to land on a spoken word, lines that draw instead of fade, small persistent detail marks, hard handoffs instead of crossfades, deliberate rhythm, and restraint to one idea per graphic. Saved as a reusable Claude skill, those rules turn the same AI and the same request into something that reads as designed rather than generated.
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.
Create a free account →Where the time goes.
00:00 – 00:52
01 · The AI-built map, no keyframes
Cold open: a described hike becomes a rendered topographic flythrough graphic that drops onto the timeline with zero manual animation, framed as proof Resolve 21.1's AI integration changed what motion graphics means.
00:52 – 01:50
02 · The default AI look, and the one missing thing
Shows the generic, default-colored AI graphic everyone gets, argues the gap isn't the tool, it's taste, and previews that the opening map render failed once and had to be thrown out.
01:50 – 02:39
03 · What you need and what it costs
Requirements: a paid AI plan that runs locally (not just chat), the free open-source HyperFrames tool, and DaVinci Resolve, free or Studio.
02:39 – 03:08
04 · Installing HyperFrames
Four-step install: open the local AI, make a project folder, paste the HyperFrames link and ask it to install, then approve the permissions it requests.
03:08 – 04:04
05 · What HyperFrames actually does
Frames HyperFrames as a camera for AI-written animations: it records a browser-based animation frame by frame into a deterministic, re-renderable 4K video file rather than gambling on a generated clip.
04:04 – 06:56
06 · The seven rules that kill the AI look
The core teaching block: one locked visual system, graphics timed to land on the spoken word, lines that draw instead of fade, a quiet detail layer, hard handoffs instead of crossfades, deliberate rhythm, and restraint to one idea per graphic, then saved as a reusable skill.
06:56 – 07:52
07 · Fixing the map with camera rules
Revisits the opening map graphic's failed first render, explains the fix was adding explicit camera rules (keep the whole route in frame, don't cover trail detail), not more style rules.
07:52 – 08:48
08 · Getting graphics into the edit
Two workflows: export and drag clips in manually with transcription, or in Resolve 21.1 Studio let the AI read the timeline directly and drop finished graphics onto a new track above an untouched duplicate of the main timeline.
08:48 – 10:07
09 · Fenn Motion Studio: the packaged version
Pitches the paid skill that bundles the rules into six named visual styles (kinetic type, data story, analog collage, keynote glass, neon trace, map journey), then demos it live on a second, unseen hike.
10:07 – 11:21
10 · Who it's for, and the close
Closes on who the paid skill solves for (generic-looking output, not knowing what to ask for, fear of setup) and reframes the whole video: the software was never the hard part, taste was.
Lines worth screenshotting.
- HyperFrames is free and open source, built by HeyGen, and works by recording an AI-written animation frame by frame into a real rendered video file.
- Unlike typical AI video generators, HyperFrames output is deterministic: rendering the same prompt twice produces the exact same result, and changing one word only changes that word.
- The same AI and the same prompt produce a generic, default-colored, same-sized-everything graphic until you give it explicit style rules, the model was never bad at animating, it just had no taste fed into it.
- Rule one is to lock a single visual system for the whole project: dark background, one accent color, off-white text, one bold font and one small label font, with zero navy, purple, emojis, made-up numbers, random gradients or blanket drop shadows.
- Rule two times every graphic to land on the specific spoken word it illustrates, using word-level transcript timing, which is what makes a graphic feel edited instead of slapped on top of narration.
- If a graphic just repeats the words already being spoken, it's a caption, not a motion graphic, the graphic has to visually show the idea instead.
- Rule three replaces fade-ins, called out as the most obviously AI-looking default move, with a bright point of light that draws each line with a glowing tail behind it.
- Rule four adds a quiet detail layer, tiny numbered labels and small corner marks that nobody consciously reads but that make a graphic subconsciously register as more expensive.
- Rule five bans crossfades between graphics in favor of a hard handoff, a line becoming the next word's underline or a card sliding out as the next slides in.
- Rule six is rhythm: hit hard on the spoken word, alternate busy and empty moments, and hold the important word on screen long enough to actually read.
- Rule seven is restraint, one idea per graphic, six words max, generous empty space, and the rule of thumb 'when in doubt, take something out.'
- Saving the seven rules as a reusable AI skill means you never retype your style, the AI applies the same taste to every new prompt automatically.
- DaVinci Resolve 21.1 Studio lets a connected AI read the editor's timeline directly and drop finished graphics onto a new track above a duplicated copy of the main timeline, so the original footage is never touched.
- Because each generated graphic lands as its own clip, you can re-render just the one that's wrong instead of regenerating the entire video.
- A topographic flythrough map graphic used real USGS elevation data and OpenStreetMap trail data rather than invented visuals, and it still took real render time (the first version took 22 minutes and was thrown away).
- The first map render failed not because of taste rules but because of unstated camera rules: the creator had to explicitly specify that the whole route stays visible in frame and the camera avoids covering trail details.
- The underlying claim is positioned as a major cost shift: hiring a motion designer or editor for custom animated graphics like these would historically run into the thousands of dollars.
AI graphics look generic for one reason: no rules, not no skill.
WHAT TO LEARN
The gap between default AI-generated graphics and designed-looking ones is a short list of explicit, repeatable style rules, not a better tool or more technical skill.
01The AI-built map, no keyframes
- A fully-rendered, data-accurate motion graphic can be produced from a single plain-English description with zero manual keyframing, if the underlying tool and prompt are set up correctly.
02The default AI look, and the one missing thing
- The default output of an AI motion-graphics tool and a rule-constrained one can come from the identical model and prompt, the only variable that changed was the instructions given.
03What you need and what it costs
- A tool that runs locally on your machine, not just inside a chat window, is a real prerequisite for AI tools that need to install software or write files directly.
- A free, open-source utility plus a free tier of your main editor is often enough to start, the paid layer is optional polish, not a requirement.
04Installing HyperFrames
- New AI-driven tools increasingly install themselves from a single pasted link and a plain-English request, the setup bottleneck has mostly moved to approving permissions, not configuring software.
05What HyperFrames actually does
- A deterministic, re-renderable output (same input always gives the same result) is a meaningfully different and more production-usable category than a one-shot AI video generator.
06The seven rules that kill the AI look
- Lock one consistent visual system before generating anything: one background tone, one accent color, one bold font and one label font, and explicitly ban the defaults that read as AI.
- Time every generated element to the exact word or beat it illustrates rather than letting it simply coexist with narration.
- Replace soft defaults like fades with a motion technique that implies intent, small deliberate choices read as designed even when they're generated.
- Restraint is a rule, not a cut corner, limiting to one idea and a tight word count per unit of output consistently outperforms cramming more in.
07Fixing the map with camera rules
- When a generated result fails for a structural reason (a bad camera path, not a style problem), diagnose and fix that specific missing rule rather than assuming the whole approach is broken.
08Getting graphics into the edit
- Keeping generated output as individual re-renderable clips on their own track, above an untouched copy of the main timeline, lets you fix one bad element without regenerating the whole project.
09Fenn Motion Studio: the packaged version
- Save your working rules as a persistent, reusable instruction set (a skill) so a tool enforces your taste automatically instead of you re-explaining it every session.
- Packaging a proven rule set into named, ready-to-use styles is a way to sell the taste you already built, not just the underlying free tool.
10Who it's for, and the close
- A commoditized AI capability still has a real cost of entry that holds value: custom motion-graphic work that used to require specialist hires is now reachable with a paid local AI plan plus a well-specified rule set.
Terms worth knowing.
- HyperFrames
- A free, open-source tool made by HeyGen that lets an AI assistant write browser-based animations and records them frame by frame into a real rendered video file.
- Claude Code / ChatGPT with a local agent
- A paid AI assistant that runs on your own computer rather than only in a chat window, required so it can install software and write files like HyperFrames needs.
- Skill (AI skill)
- A saved set of instructions an AI assistant reads automatically on future requests, used here to store the seven style rules so they don't need to be retyped each time.
- Fusion composition
- DaVinci Resolve's native node-based motion graphics format; HyperFrames can optionally save its output as an editable Fusion composition directly on the timeline instead of just a flat video file.
- GPX file
- A GPS data file format storing a recorded route's coordinates and elevation, usable as a real data source for generating an accurate map or journey graphic.
Things they pointed at.
07:30toolUSGS elevation data
07:30toolOpenStreetMap trail data
Lines you could clip.
00:00
“I described it, AI built it, and it landed right there on its own track in my DaVinci Resolve timeline.”
01:33
“It's rules. The AI was never bad at animating. It just didn't have any taste because you didn't give it any.”
05:22
“A fade in is like the most AI thing you could ever do.”
06:08
“When in doubt, just take something out.”
11:26
“The software was never the hard part. Keyframes was really never the hard part. But taste was.”
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.
metaphoranalogy
That's a 35 mile hike from Aspen, where I grew up over a 12 ,500 foot pass into Crested Butte. Real elevation data, the actual trail. And I didn't set one keyframe.
I described it, AI built it, and it landed right there on its own track in my DaVinci Resolve timeline. Type that rises on the word. Lines drawn by light.
Cards that stack up the second I list them. Every one of these was just one sentence. Graphics like this used to mean After Effects or spending hours in Fusion and playing with nodes.
But Resolve 21 .1 lets an AI live inside your timeline, and that quietly flipped the whole thing. Motion graphics stopped being something you animate, and now it's something you describe. Now, if you've tried this, you've probably gotten something like this.
Default colors, generic layout. It looks like every AI graphic ever made. And that's not AI being dumb.
It's just missing one thing. And that one thing is the whole reason why mine look like this instead of that. So I'm going to walk you through the whole thing step by step.
It doesn't matter if you use Claude or ChatGPT, and it doesn't matter if you've used AI for anything but chatting. We'll set it up, see what it makes on its own, and then fix the AI look. And stick around, because remember that map at the start?
The first version took 22 minutes to render, and there's a reason why I had to throw that one away. All right, let's start with what you need and what it's going to cost so nobody's surprised. First, you need an AI that works on your computer, not just a chat.
That part is paid. You'll need a paid plan no matter which AI that you pick. Second is Hyperframes.
It's free and open source, and it's made by HeyGen. Third is DaVinci Resolve, or a video editor of your choice. The free version does work fine, but the studio version has a little bonus that I'll show you later.
The AI can put the tracks on the timeline for you. All the links to everything I just said are linked down in the description. Here's the part that kind of trips people up.
Step one, just open the AI on your computer. Step two, I made a new folder and I just called it motion graphics. That's where all of your work is going to be saved.
Step three, copy the hyperframes link down in the description, put it into your chat box and just say, Hey, install this and just make sure to include anything that you think would be useful with this install. Step four, it most likely is going to ask you if it's okay to do certain things, which you will have to allow it to do.
So what did we actually just install? . Your AI will write the animations kind of like a little webpage that moves.
HyperFrames records it frame by frame and hands you a real video file. And that's kind of like the secret weapon here because this isn't like all those other AI video generators out there where you type something in and kind of like roll the dice, you know? The text is real.
spelled right and in 4k if you try to render the same thing twice it will render exactly the same if you want to change a word it will just change that word and keep everything exactly the same with that said if you don't want just video files if you specifically ask it to put the actual compositions and the fusion compositions on your timeline it will save them there for you so it's so let's try it just just using plain english Use HyperFrames to make a 10 -second motion graphic listing the five things in my editing kit.
Camera, laptop, editing app, sound library, and my favorite plugin, which is Afterlight. It's a new plugin that I've been working on that's going to be insane. Render it as a video.
All I have to do is then just drag it into my Dimension Resolve timeline, and boom, it plays back. That's HyperFrames on its own. One sentence in, one video out.
Not long ago, this was like an entire motion graphics designer project. But look at it. But like really look at this.
You've probably seen this graphic a thousand times. Default colors, everything's kind of the same size. Nothing's really telling your eye where to go.
And it looks AI made because it is AI made. And here's the thing. Mine are exactly the same.
So what's actually different? It's rules. The AI was never bad at animating.
It just didn't have any taste because you didn't give it any. So here are the seven rules that I gave mine. These are also written down in the description so you can copy and paste them into your own prompts.
Rule one, pick one look and never break it. Dark background, one accent color, off -white text, one big bold font, and one small font for little labels. Glow only goes on the accent.
We don't want any navy or purple background, any emojis, any made -up numbers, random gradients, or drop shadows on everything. That looks like AI. Kill all of it.
Rule number two, every graphic lands on a word. This is the one that makes it feel edited rather than just kind of slapped on. For a real video, the AI will read what you said with the timing of every word before it builds it.
It hands you the plan. This line, this graphic, this moment. You say yes, and it just builds it.
And the graphic should show the idea, not just repeat it. A number becomes a counter. A list becomes a stack as I say each thing.
If it's just your words typed on a screen, that's just a caption, not a motion graphic. Rule number three, lines get drawn. They don't fade in.
. Instead, a bright point of light draws the line with a glowing tail behind it. That's the difference between a line appearing and a line being made.
Rule number four, the detail layer. Tiny labels, little numbers like 01 and 02, small marks on the corners. Nobody consciously reads them, but your brain notices when they're missing.
In some cases, it's what makes a graphic look kind of expensive. If you use those, keep them small and quiet. Rule number five, and one of my personal favorites, hand off, do not crossfade.
Every graphic should pass the baton onto the next. A line becomes the next word's underline. A card slides out and the next one comes in.
Never just a dissolve. Rule six, rhythm. This is where being an editor really helps.
Hit hard on the word. Busy, then empty, then busy again. And hold the important word on the screen long enough for people to be able to read it.
Rule number seven, restraint. Just do one idea at a time. Three to six words max.
Lots of empty space. And my favorite rule of all, when in doubt, just take something out. Now watch what telling it these rules does.
Same list, same AI. I just paste the rules into my message. Same AI, same list, already better.
Just better instructions. Already way better, right? And this is mine.
Same list, just one more time. The difference, I don't paste the rules in every single time. I saved them as a skill.
A skill is basically a recipe card for your AI to read every single time. So you never really have to explain your style again for that type of motion graphic. And mine have a lot more than seven rules in it.
So that's the taste. And that's actually where my map in the very beginning fell apart. Remember that fly through at the start?
It took me 22 minutes to render and I watched it back and I threw it out. The overall graphic and the concept worked really well. The intro was a bird's eye view.
And then I wanted it to kind of transition into more of like a tight following the trail type of view. But the problem was, was the transition between the two. And also the camera was really shaky.
It was like following the actual little dot as it went through the trail. I didn't want that. So what fixed that was giving the camera rules too.
The whole route was visible within the first frame. I had to make sure the camera never covered the trail wiggles. So it just glided over the...
through the valley and what's crazy is none of it was made up the elevation is real usgs data the trail comes from open street map and it's also credited right there on the screen the one honest thing though it did take a while to render even the second time this is an instant 4k graphics do take some time to render think of it as like a collaborator not necessarily just like a magical button At least for now.
I approve the plan, but I also have to watch every clip. So how do you get all of this into your edit? Well, there's two ways.
The first way is just export the video out of whatever video editor that you're using and then just add the file into the chat. It transcribes it right there on your computer so it knows every single word that you say. Then it gives you a graphic with a time code and you have to drag them in by hand.
Here's the bonus way if you're using DaVinci Resolve 21 .1 Studio. Go to file, then set up AI assistance and connect your AI. Now, your AI can read your timeline directly, so skip the exporting and dragging.
Now, at the end of my prompt, I can say, put these on my timeline. It makes a copy of my timeline first, which is super nice because I don't want anything to be messed up on my main timeline. And then it drops the graphic above on a new track.
Each one right on the words when I'm talking. And because each one is its own clip, I can choose whether I like one or not, and then I can just re -render the one that I need to work on instead of the entire video file. And yeah, every single graphic you've seen in this entire video was made with this workflow.
So that's the whole workflow. Hyperframes is free. The rules are in the description.
And you can build every bit of this yourself. Honestly, I want you to. But the taste, the styles, all the stuff that I learned throwing out version one of everything, that's what took time.
And if you'd rather skip that part, I packaged my exact setup. It's called Fenn Motion Studio. It's a skill.
So you install it once in Clod or Chachi PT, and it knows six complete styles. Kinetic type. Huge type that rises in word by word.
That's your hooks. Data story. Dark, glossy cards for list and numbers.
Analog collage. Torn paper, tape, red marker for opinions and hot takes. Keynote glass.
A bright launch event look with frosted glass and a cursor that clicks. Features, software, sponsor reads. Neon trace.
Circuits drawn by light. Tech stuff. Show real moments.
And map journey. the one you saw at the start, any route, anywhere, or use your own GPX file. But don't take my word for it.
Let's just build one right now, somewhere that isn't from Aspen to Crescent Butte. I did a hike in Utah called Devil's Garden, and it was amazing. The arches were incredible.
The overall trail length was perfect. It wasn't too hard. And yeah, it was just...
Such a memorable moment in my life. So what you just saw, I've never made a map of that before. That's the first time anyone's seen it, including me.
So who's this for? If your graphics keep coming out looking like AI, every rule I showed you, and a lot more, is baked into six styles. If you sit there not knowing what to ask for, there are 12 copy and paste prompts.
If you're worried about it not landing on what you're saying, it plans off of your words. So every graphic lands on a word. And if the setup process is scary for you, you paste one prompt.
the skill, checks your tools, and makes a test render so you know it's working. There's a step -by -step install guide too. So basically, motion graphics that look like a designer made them landing on your words right in your edit and you never had to open after effects or fusion honestly this is a huge jump from what people had to pay just a couple months ago because if you had to hire a motion graphic designer or an editor and you wanted to make some custom motion graphic effects like this it would cost you thousands of dollars so again it's called fen motion studio it's the first link down in the description because here's what i found out making all of this the software was never the hard part keyframes was really never the hard part but taste was.
And now, you've got mine. Thanks for watching. See you guys in my next video.
Peace.
The Hook
The bait, then the rug-pull.
He describes a 35-mile mountain hike in one sentence and watches an AI render it as a real, data-accurate topographic flythrough, landing it straight onto his DaVinci Resolve timeline without a single keyframe set by hand.
Named ideas worth stealing.
04:04list
The Seven Rules (anti-AI-look system)
- One look, never broken (dark bg, one accent color, off-white text, one bold font, one small label font)
- Every graphic lands on a word (timed to the exact spoken word it illustrates)
- Lines get drawn, not faded in (a glowing point draws the line with a tail)
- The detail layer (tiny numbers, corner marks, kept small and quiet)
- Hand off, don't crossfade (each graphic passes the baton to the next)
- Rhythm (hit hard on the word, alternate busy and empty, hold key words)
- Restraint (one idea, six words max, lots of empty space)
A seven-rule prompt system the creator pastes into every AI motion-graphics request, and later saves as a persistent skill, that converts default AI output into graphics that look intentionally designed.
Steal forany AI-assisted graphic, caption, or UI generation prompt where the default output reads as generic
How they asked for the click.
VERBAL ASK
08:48product
“It's called Fenn Motion Studio... you install it once in Claude or ChatGPT, and it knows six complete styles.”
Soft-sells a paid Claude/ChatGPT skill that packages the free workflow's taste rules into six ready styles, positioned after the full free method has already been taught in detail, then closed again at the very end of the video.
MENTIONED ON CAMERA
FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
OTHER LINKSAlso linked in the description.
Visual structure at a glance.
