Modern Creator
Andy Lo · YouTube

Remotion vs HyperFrames: Which AI Video Tool Wins?

A controlled 12-minute experiment: same AI, same storyboard, two frameworks — to find out where each one is actually stronger.

Posted
6 days ago
Duration
Format
Tutorial
educational
Views
2.6K
71 likes
Big Idea

The argument in one line.

When every variable except the framework is held constant, Remotion and HyperFrames expose their true identities: one is a developer-controlled React render pipeline, the other is an AI-directed HTML-to-video system that trades frame precision for creative headroom.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…
  • You use Claude Code, Codex, or Cursor to build video content and want to know which rendering framework to route your prompts through.
  • You have tried Remotion and are wondering whether HyperFrames solves a genuinely different problem or is just a shinier wrapper.
  • You want a methodologically honest comparison — controlled inputs, not cherry-picked demos — before committing to a video production stack.
SKIP IF…
  • You are looking for a no-code video editor; both tools require an AI coding agent or developer to operate.
  • You want a large-sample production verdict; this is a single-project controlled test with one creative brief.
TL;DR

The full version, fast.

Both frameworks can produce polished cinematic explainers from the same brief and the same AI model. The real difference is architectural: Remotion treats every frame as a React render, giving developers precise control over timing, composition, and output fidelity. HyperFrames has AI agents generate HTML, CSS, and JavaScript, then renders that to deterministic video — freeing the AI to spend more of its context on visual creativity rather than structural correctness. Neither is universally better; they are optimized for different jobs and different builders.

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 →
Chapters

Where the time goes.

00:0001:35

01 · Cold open — the experiment thesis

Introduces the head-to-head premise: same AI (Claude Opus 4.8), same storyboard, same three build prompts, two frameworks. Promises a side-by-side production result.

01:3602:40

02 · HyperFrames showcase review

Reviews HeyGen public showcase videos to identify recurring visual production patterns across projects with different stories and styles.

02:4103:49

03 · Source code pattern mining

Digs into the public HyperFrames repo to reverse-engineer effects (glassmorphism, RGB splits, shader effects, portal reveals) and packages them into a reusable production skill.

03:5005:03

04 · Project setup for HyperFrames

Walks through the three required inputs: storyboard and direction document, custom HyperFrames production skill, and phased build prompts.

05:0408:32

05 · Building the HyperFrames version

Three-phase Claude Code build. Shows Hyperframes Studio output after each phase. Final result: smooth motion, cohesive transitions, clean typography.

08:3310:44

06 · Setting up and building the Remotion version

Repeats the exact same experiment with the official Remotion skill replacing the custom HyperFrames skill. Same three-phase build, same AI model.

10:4512:38

07 · Final thoughts and conclusion

Explains the architectural difference and delivers the verdict: HyperFrames = AI creative headroom; Remotion = developer precision and pipeline ownership.

Atomic Insights

Lines worth screenshotting.

  • Remotion and HyperFrames are not competitors — they are optimized for different builders: one for developers who want frame-level control, one for AI agents that want creative headroom.
  • Packaging production patterns into a reusable skill file lets Claude build on proven techniques instead of rediscovering them from scratch on every project.
  • Breaking an AI video build into three phases (foundation, scenes, polish) produces more reliable results than trying to generate everything in a single prompt.
  • The most honest framework comparison controls every variable except the framework itself — same AI, same storyboard, same prompts.
  • HyperFrames trades frame-level precision for creative flexibility; Remotion trades creative flexibility for structural correctness and pipeline ownership.
  • Studying the source code of public showcase projects is faster and more actionable than studying the finished videos alone.
  • A custom production skill extracted from real launch videos gives an AI model a stronger foundation than a generic system prompt.
  • A storyboard that defines how a video should feel, not just what it should show, produces more consistent AI-generated output across multiple builds.
Takeaway

Two frameworks, two different definitions of control.

WHAT TO LEARN

The framework you choose for AI video production is really a choice about where you want control to live — in the developer or in the AI agent.

  • Remotion is a React application where every frame is authored in code — this gives developers exact control over timing, composition, and output fidelity, but it means the AI spends more of its context budget on structural correctness rather than visual creativity.
  • HyperFrames has AI agents generate HTML, CSS, and JavaScript, then renders that to deterministic video — this frees the AI to invest more effort in motion quality, transitions, and visual polish at the cost of frame-level precision.
  • Packaging production patterns from real reference projects into a skill file before building produces more consistent AI output than starting from a blank prompt every time.
  • Breaking a complex AI video build into three sequential phases (structure, scenes, polish) reduces failure modes by narrowing the scope of each generation step and giving you a visible checkpoint before committing to the next.
  • The most honest way to compare two tools is to control every other variable — same AI, same creative brief, same prompts — so that the framework is the only thing that can explain the differences in the result.
Glossary

Terms worth knowing.

HyperFrames
An HTML-to-video framework by HeyGen built for AI coding agents. The AI generates HTML, CSS, and JavaScript; HyperFrames renders that output into deterministic video files.
Remotion
A React-based video framework where every animation and visual element is expressed as code and rendered frame by frame. Gives developers precise control over the entire pipeline.
Production skill
A packaged context file fed to an AI coding agent before a build, containing reusable production patterns (effects, animation systems, transition styles) extracted from existing projects so the AI does not have to rediscover them.
Deterministic video output
A render that produces the same frames every time, regardless of when or how many times it is run — as opposed to generative video that may vary between runs.
Glassmorphism
A UI design trend using frosted-glass visual effects: semi-transparent backgrounds, background blur, and subtle borders to create a layered depth effect.
Chromatic RGB split
A visual effect that separates the red, green, and blue color channels of an image or text element and offsets them slightly, creating a glitch or prismatic fringe appearance.
Resources

Things they pointed at.

Quotables

Lines you could clip.

03:21
Instead of asking Claude to reinvent these techniques every single time we start a project, why not just package them into a reusable production skill?
Clean standalone insight about AI workflow efficiency — no setup neededTikTok hook↗ Tweet quote
04:59
We are starting from the exact same foundation, and watching both build a video from there.
Tight premise setup — good pull-quote for a comparison threadnewsletter pull-quote↗ Tweet quote
11:54
HyperFrames let the AI spend more effort on creativity and polish, while Remotion spends more effort on precision, structure, and correctness.
The cleanest one-sentence verdict in the entire videoIG reel cold open↗ 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.

analogy
00:00Just a few months ago, if you wanted AI to build a cinematic explainer video from code, there was really one name everyone talked about, Remotion. Then HeyGen released Hyperframes, an HTML to video framework built specifically for AI coding agents like low-code, Codex, and Cursor. So, instead of using React components yourself, you can now direct an AI using plain language and let it generate the production system for you.
00:25So, that raises an interesting question. If you give the exact same AI the exact same storyboard and the exact same creative brief to both tools, which one produces the better video? And today, in this video, we are going to find out by building the exact same cinematic explainer twice, once in Remotion, once in Hyperframe.
00:44And even if you are not a developer, the results might surprise you. So, in order to compare Hyperframes and Remotion fairly, we need to control as many variables as possible. So, here is the experiment.
00:56Step one, use the exact same storyboard, creative direction, and production requirements. Step two, use the exact same AI model, in our case, Claude Opus 4.8.
01:05And step three, build the entire video twice, once using Remotion, once using Hyperframes, and then compare the results side by side. And today, we are building exactly that, two complete cinematic explainer videos, roughly 80 seconds each, created from the same three prompts, the same AI, and the same creative brief. The only thing changing is the framework.
01:27And by the end of this video, you will know where Remotion performs better, where Hyperframes takes the lead, and which one makes the most sense for your workflow. And before we start building, let's just take a look at some of the showcase projects published by the Hyperframes team.
01:43So, looking through these projects, one thing becomes obvious very quickly. Like, they are all solving different problems, but many of them are actually using similar production techniques underneath. And for this launch video, for example, combines motion graphics, custom typography, visual effects, transitions, voice over, and multiple animation system into one single production.
02:05And then, when you move to the website to Hyper Frames demo, and even though the visual style changes completely, many of the same design principles are still there. And the same thing happens again with the timeline editor launch, a texture launch, and the VFX showcase.
02:21So, different stories, different visuals, but a lot of recurring patterns start appearing once you have watched enough of them. So, now, from a viewer's perspective, you can just see the finished video, but from a builder's perspective, the interesting part is figuring out how these effects are actually created. Unfortunately, HeyGen made the source code for these projects publicly available.
02:45So, let's jump into a view of the repo and take a look at them. Even though you're not a developer, don't worry too much about like what you're seeing on screen. Because the important thing is not the code itself anymore.
02:56What we are actually looking for are the patterns that can keep showing up across multiple projects. And as we started going through these repos, we kept noticing the same techniques appearing again and again. And things like glassmorphism, custom cursor interactions, typewriter effects, texture masked typography, chromatic RGB splits, shader effects, portal reveals, and a very specific transitional style that appears throughout many of the launches.
03:26So, instead of asking Claude to reinvent these techniques every single time we start a project, why don't we just package them into a reusable production skill? So, that's exactly what we did. This custom Hyper Frames skill is essentially a collection of production patterns extracted from real Hyper Frames launch videos.
03:44And instead of teaching Claude a single project, we are actually teaching it the techniques behind all those projects. And that means when we start building later, Claude already understands many of the animation systems, visual effects, motion patterns, and production conventions used throughout these showcase videos. Now, you may ask, "Hey Andy, does that guarantee a great result?" Not exactly, because you still need a good storyboard, a good creative direction, and good prompts.
04:12But, it does give Claude a much stronger foundation to work from. And that is important because for this comparison, we want both HyperFrames and Runway Motion operating at their best.
04:22And now that we have a sense of the quality bar that we are aiming for, let's set up the project and make sure that this comparison is actually fair. So, for this project, the goal here is very simple. We want to isolate the framework itself.
04:36Because in that way, when we compare the final videos later, we can be much more confident about where those differences came from. And this is actually one of the most interesting parts of the experiment, because most comparisons happen after the fact, and people look at two finished videos and try to guess why one looks different from another.
04:56And we are actually doing the opposite. We are actually starting from the exact same foundation, and watching both built a fourth from there. Here we are in Claude Code, and before we can start building, we need three things first.
05:09The first one is the storyboard and direction document. You can think of this as the creative blueprint for the entire project, because it defines what the video is trying to teach, the story we want to tell, the visual style, the motion philosophy, and the structure of every single scene. And if we scroll through it, you can see that we are not just describing the video, we are actually defining how it should feel.
05:34For example, the core idea behind this project is that Runway Motion and HyperFrames solve the same problem from two different directions. One is developer first, the other is AI first. And instead of simply, like, talking about the difference, the video itself is designed to demonstrate it.
05:52And then, next, we define things like typography, color palettes, motion rules, transition styles, and the individual scenes that make up the final video. And we do not need to obsess over every single detail here, because the important thing is giving Claude a very clear creative direction before we start building. And the second thing we need is our custom hybrid frames production skill.
06:16And we already talked about this earlier, but basically, this is a collection of production patterns extracted from real hybrid frames launch videos. Things like glass morphism, custom transactions, shader effects, RGB splits, portal refuse, typography treatments, animation systems, and other techniques that repeatedly showed up across the showcase projects.
06:39So, instead of forcing Claude to rediscover those techniques every single time we start a project, packaged them into one reusable skill. So, Claude already understands many of the production patterns we want it to use.
06:54And finally, we have our built prompts. And the build process is split into three phases. The first prompt builds the foundation and project structure.
07:03The second prompt builds the actual scenes and effect. And the third prompt focuses on polish and final export. So, breaking the build into phases tend to produce much more reliable results than trying to generate everything in one single prompt.
07:19Everything is already prepared, and we have the storyboard and direction, we have our custom hybrid frames production skill, and we have Claude Opus 4.8 ready to do the work. So, the only thing left is to see how hybrid frames approaches the build process. So, let's begin with prompt one.
07:36And I'll copy it over and paste it into Claude code and hit enter. And now, this is going to take a few minutes, so let's just wait for Claude to finish. So, prompt one is done and at this stage we are not expecting this finished video yet and we are actually just checking whether the foundations are in place.
07:53So, let's open Hyperframes Studio and take a look. You can see that the project structure has been created, the scenes skeletons are there and the overall framework looks pretty healthy. Like nothing fancy yet, but everything looks good.
08:05So, let's move on to prompt two. So, just like before, I'll just paste it in send off and let Claude do its thing. So, now prompt two is complete as well.
08:15For the sake of time, let's jump straight into the final phase. We'll take prompt three, paste it into Claude code and let it handle the final polish pass. That has finished rendering and let's jump into the final preview and see what we got.
08:29And honestly, this came out pretty well. Like the motion feels smooth, the transitions are cohesive, the typography is clean and most importantly, it follows the creative direction we defined at the beginning. So, considering this was built almost entirely through AI assisted production, the result is surprisingly polished.
08:49So, now the Hyperframes version is complete, it's time to repeat the exact same experiment inside Remotion. And again, nothing else is changing. We are actually using the same storyboard, the same creative direction, same AI model and the same three built prompts as well.
09:05And the only thing changing is just the framework. And this is exactly what makes this comparison interesting, right? And here we are back in Claude code.
09:13And if we look at our project files, you can notice that most of them are exactly the same as before. The only thing missing here is the custom Hyperframes production skill. And instead of using our custom skill, we are now using the official Remotion skill, which is designed around React frame-based animations and Remotion's architecture.
09:33So, everything else stays as close as possible to the original experiment. So, let's begin with prompt one. I'll just copy it over and paste it into Claude code and hit enter.
09:43And this is going to take a few minutes, so let's just wait for Claude to finish. All right, prompt one is done. And just like the Hyper Frames version, we are not expecting a finished video yet.
09:54At this stage, we are actually just checking whether the foundations are in place. So, let's just open the project and take a look. So, you can see that Claude has created the comment section structure, the scene skeletons, and the overall project architecture.
10:08And again, nothing is exciting yet. We're simply making sure the groundwork is solid before we start building the actual scenes. So, everything looks good.
10:16Let's just move on to prompt two. I'll just paste it to Claude code, send it off, and let Claude continue to build. And now, prompt two is complete as well.
10:25And at this point, most of the video has been generated. The scenes will be there, the animations, and the over structure is starting to resemble the final product. But just like before, let's go ahead and run the final phase.
10:38So, we just take prompt three, paste it into Claude code, and let it handle the final polish pass. All right, that has finished rendering. And now, let's jump into the final preview and see what we've got.
10:49And honestly, this came out surprisingly well, too. Like the motion feels polished, the facial quality strong, and most importantly, it stays pretty close to the creative direction we defined at the beginning. And what's interesting is that even though the final video is trying to tell the exact same story, you can already start differences in how the framework approaches the problem.
11:09So, now that both versions are complete, let's wrap everything up and talk about the takeaways. All right, so here's what the experiment actually delivers.
11:17The same AI, the same prompt, the same storyboard. Two complete production-ready videos built using two completely different approaches. Remotion treats video as a React application.
11:28Every animation, transaction, and visual element is expressed through code and rendered frame by frame. And hyperframes take a different approach. AI agents generate HTML, CSS, and JavaScript, and hyperframes renders the result into deterministic video output.
11:45And after building both, the takeaway is pretty simple. The AI assists in both, actually. Hyperframes let the AI spend more effort on creativity and polish, while Remotion spends more effort on precision, structure, and correctness.
11:59So, they are actually optimized for different jobs. So, remember to try this out yourself and build the same project in both frameworks, so that you can learn more in a weekend than you will from weeks of just reading opinions online. If there are other AI video frameworks that you want us to compare, let us know in the comment section below.
12:17And if you want more in-depth tutorial like this, feel free to join our Any No-Code Community. You can find the links in our description below. And in our community, you can find exclusive prompts and tutorials that can help you make more money with AI.
12:29And as always, if you found this video helpful, drop a thumbs up and hit the subscribe button for more videos like this in the future. I'll see you in our next one.
The Hook

The bait, then the rug-pull.

Two AI video frameworks walk into the same storyboard. One is developer-first, built on React and frame-level precision. The other is AI-first, built on HTML-to-video rendering and creative flexibility. The only way to know which one is actually better is to give them both the exact same brief and watch.

Frameworks

Named ideas worth stealing.

03:25concept

HyperFrames Production Skill

A packaged context file of production patterns extracted from real HeyGen launch videos and fed to Claude before a build. Covers glassmorphism, chromatic RGB splits, shader effects, portal reveals, typewriter effects, texture-masked typography, and animation systems.

Steal forAny AI-assisted video project — extract techniques from reference videos and package them into a skill file so Claude starts with a proven foundation
06:56model

Three-phase build prompts

  1. Phase 1: Foundation and project structure
  2. Phase 2: Scenes and effects
  3. Phase 3: Polish and final export

Breaking AI video generation into three sequential prompts rather than one. Each phase has a clear scope and a visible check before moving on.

Steal forAny multi-scene AI video build — reduces hallucination and structural errors by narrowing the scope of each generation step
CTA Breakdown

How they asked for the click.

VERBAL ASK
12:28subscribe
If you found this video helpful, drop a thumbs up and hit the subscribe button for more videos like this in the future.

Standard end-card ask. Also promotes AndyNoCode Skool community for exclusive prompts and tutorials.

FROM THE DESCRIPTION
PRIMARY CTAWhere the creator wants you to go next.
AFFILIATECommission earned if you click.
OTHER LINKSAlso linked in the description.
Storyboard

Visual structure at a glance.

host intro
hookhost intro00:00
Remotion logo + VIDEO IS CODE
contextRemotion logo + VIDEO IS CODE00:37
HyperFrames showcase
contextHyperFrames showcase01:46
GitHub repo
valueGitHub repo02:41
storyboard doc in VS Code
valuestoryboard doc in VS Code03:50
HyperFrames build phase 1 in Claude Code
valueHyperFrames build phase 1 in Claude Code07:04
Remotion build setup in Claude Code
valueRemotion build setup in Claude Code08:33
Remotion is React explainer slide
valueRemotion is React explainer slide11:03
REMOTION / REACT APPLICATION title card
valueREMOTION / REACT APPLICATION title card11:27
closing host
ctaclosing host12:38
Frame Gallery

Visual moments.

Chat about this