Building Tony Stark's 'Jarvis' with Claude Fable 5

DesignCourse| 00:06:14|Jun 17, 2026
Chapters9
Introduction to the AI assistant and the goal to automate studio management, asset generation, and marketing tasks.

Tony Stark vibes meet real-world AI rigs: DesignCourse's creator builds a Claude Fable 5-powered personal assistant to control a studio, manage projects, and generate live light shows.

Summary

DesignCourse's latest experiment turns a home studio into a high-tech AI playground. The creator showcases a Tony Stark–style personal assistant built with Claude Fable 5, OpenAI real-time GPT-2, Grok, and a mix of local and cloud models. He demonstrates how the AI can manage studio equipment, pull YouTube channel analytics, run marketing campaigns, and generate on-brand assets. A key feature is a 24-foot projection wall where a 3JS scene changes in response to the AI’s spoken outputs. The system is wired to control lights, DMX lighting, projector walls, and music lighting effects, illustrating how automation can scale from creative ideas to live production. While the tech looks impressive, he candidly notes the current fragility of tool calls and the need for careful system prompts to ensure reliable behavior. The long-term dream is an almost hands-free workflow: wake up, coffee on, and a fully connected smart studio that provides business analytics and an immersive lighting show. He teases future capabilities like door-entry triggers, live camera feeds, and automated consistency across multiple projects. Throughout, he emphasizes the challenge of monetizing marketing in the AI era, choosing a practical path that blends automation with real-world business goals. Viewers are invited to follow the ongoing project progress and subscribe for more updates.

Key Takeaways

  • Claude Fable 5 can drive a multi-tool studio rig, including lights, DMX, projectors, and music-driven shows.
  • Cloud and Grok AI voices offer different strengths; cloud-based GPT-2 provides clearer output while Grok supports live tool integrations.
  • The system can pull real-time YouTube analytics to surface top-performing videos and generate on-brand assets.
  • Effective AI automation hinges on well-crafted system prompts and modular access to tools for predictable results.
  • The long-term vision is a seamless, phone-to-studio workflow with automatic scene setup, occupancy detection, and business analytics dashboards.
  • Generating a music-responsive light show from a YouTube video demonstrates practical cross-domain AI capabilities for content creators.
  • The core challenge remains monetization and marketing focus, not just building AI tools.

Who Is This For?

Essential viewing for creative technologists and YouTubers who want to see how to wire Claude Fable 5 into a production studio to automate marketing, asset creation, and live lighting.

Notable Quotes

"This is my own Tony Stark AI personal assistant thing that I've been working on."
The speaker introduces the project and its aspirational vibe.
"I track your projects and goals, pull YouTube and business stats, generate on-brand images and videos, run marketing campaigns and content planning, and control the entire studio rig, lights, projector wall, streaming, and synced music light shows."
Showcases the capabilities of the AI toolkit across both business and studio control.
"Turn all of the stage lighting and the Hue lights blue. Done. Blue stage glow, baby."
Demonstrates live control of lighting through the AI system.
"The ultimate goal for me is like I'm going to wake up, I'll be able to look at my phone and press an on button, I'll get my coffee, and while that's happening, it's going to turn on all my stuff in the studio."
Outlines the envisioned seamless, automated workflow.
"There's not as much interest in [OpenClaw] it fell off the cliff, but it's not because it's not capable. It's because most people don't have a really great use case for it."
Addresses community reception and practical challenges.

Questions This Video Answers

  • How can Claude Fable 5 be integrated with a DIY studio rig for automated lighting and visuals?
  • What are practical steps to build a voice-controlled AI studio like Tony Stark's in 2024?
  • What are the differences between Cloud GPT-2 and Grok for live tool integration?
  • How to ensure reliable automation when AI tool calls fail during live production?
  • What does it take to monetize AI-driven marketing and production workflows for creators?
Claude Fable 5OpenAI GPT-2Grok AIDMX lighting3JS visualizationlive automationstudio rigAI-powered marketingContent creation workflowRiffscape
Full Transcript
There. The rig is No, come on. Turn them the [ __ ] on. This is not working right. Got it. I'll blast the rig to full [music] intensity Okay, so here, check it out. This is my own Tony Stark AI personal assistant thing that I've been working on. And really, I worked mainly when Fable 5 was out. You know, I never got onto the open claw Hermes nano claw, you know, bandwagon. But Fable 5 was so like good that I saw I was like, you know, we just built our own from scratch essentially. [music] And so, this is what it looks like. In this video, I'm going to show you the capabilities of this thing. Mind you, it's only a couple days in, um but it could still really do [music] some cool stuff. It has full control of my my light studio. But more importantly, it has access to my businesses. And the goal here is to get this thing so good that it can really just help manage my projects. It can generate assets. It can run ad campaigns. It can analyze them. It can do all that stuff. Because let's face it, building is easy now, marketing is not. So, I'm just going to show you kind of how this works. So, in this UI here, um I'll talk first about like the the actual voice aspect. Um and that is I have a cloud version, which is Open AI's real-time GPT-2. Uh and then I have local, which is actually a free version, but it sounds like crap. And then there's also Grok. Grok has a real-time API as well. And so, if I choose cloud here and I ask it a specific question, I'll turn on voice. Hey, can you tell me what my most popular video in the last 30 days was on the Design Course YouTube channel? Let me check your channel stats and pull the standout upload from the last month. Your top video was the Claude Fable 5 UI and UX one shots. It absolutely dwarfed everything else this month by a mile. Okay, so as you can see, it's able to pull it up. I've orchestrated, um you know, this this is a projection screen on a 24-ft wall behind me. Um I've orchestrated a 3JS scene where we have this kind of head model that has interesting shaders on it, and then when it talks and does stuff, it shapes the UI differently back there, which is really fun to, you know, kind of tinker around and play with. Now, I'll switch to Grok right here. And I'll ask it a question. All right, tell me what your capabilities and tools are here in this project. I've got the full live AI toolkit wired in. I track your projects and goals, pull YouTube and business stats, generate on-brand images and videos, run marketing campaigns and content planning, and control the entire studio rig, lights, projector wall, streaming, and synced music light shows. I All right, so as you can see, it it has a different voice. There's different voices you can choose from. It's not quite as clear and pristine as the Open AIs, um but it's still pretty good. All right, turn all of the stage lighting and the Hue lights blue. Done. Blue stage glow, baby. No, yeah, you actually need to turn on the actual DMX lights. Take them all blue, not just the Hue lights. No, come on, turn them the [ __ ] on. This is not working right. There's all the almost of the lights are still turned off. Make sure the brightness is 100%. intensity and reset the look. All right, finally. So, as you can see, it's kind of frustrating when it comes to the tool calling because it doesn't always do what you expect it to do. And so, that's just a matter of refining the system prompts and making sure that it has access to the different iterations based on what your intent is so that it accurately can, you know, convey and do what you want it to do. So, the ultimate goal for me is like I'm going to wake up, I'll be able to look at my phone and press an on button, I'll get my coffee, and while that's happening, it's going to turn on, you you all my stuff in the studio. My studio is like 20 ft away from the house. And when I walk in that door, I'm going to have uh cameras that are always on that will detect when a person enters the door. It'll fire that as a signal to the AI. I'll be able to see my little AI companion on the screen and it'll give me a breakdown of the stats across my businesses, what has happened in terms of traffic, what has happened in terms of money, etc. etc. etc. Now, one really cool thing I can have it do at this point is I can tell it to generate a light show for anything on YouTube. So, any track. Now, I don't want to do any track because I don't want to get demonetized, but I do have my own music and I'm going to see if it works for my own music. I want you to build a light show for my music video on YouTube. The YouTube channel name is called Riffscape. That's r i f f s c a p e and the song is called Primitive. Go. Nice. I'm queuing up the track and building a sync show for it now. So, what it's doing is trying to find the video on the YouTube API and then it's going to analyze the video and then create a light show that is in sync with the BPM and the def essentially the mood of the video. It's like it's working. All right. So, I'm excited to keep on sharing more as this, you know, project develops. And the big question is is, you you know, a lot of people OpenClaw really fell off the cliff. There's not as much interest in it. And And it's it's not because it's not capable. It's because most people don't have a really great use case for it. But I have three or four different projects that, you know, we can really test this on because that's the biggest issue right now with the vibe coders. It's like everybody can build, but nobody's making money because the hardest part was never building. Even before AI, the hardest part is is finding is marketing. I mean, that's the hardest part. So that's going to be my goal with this. Make sure to subscribe and follow up with the project the progress of this project and obviously many others. I will see you all very soon. Goodbye.

Get daily recaps from
DesignCourse

AI-powered summaries delivered to your inbox. Save hours every week while staying fully informed.