Learn robotics as you build. Course includes hours of detailed video instruction on 3d printing, assembling, wiring and programming your robot head with a detailed BOM for parts and materials. The course will guide you through the completion of the robot head ALAN, but the online store will include many new modular parts like heads, necks and shoulders to modify your robot into new characters. One platform unlimited robots.
Parts & Materials
You will build using proven controllers like the EZB, IO Tiny, Arduino and ESP 32s as controllers for servo motions LEDs and camera/vision.
You'll need various tools including a 3D printer to produce the plastic parts, use of soldering iron, heat sinks, standard tools like wrenches etc. If you are a DIY builder you will have alot of the tools and parts needed.
Discover more robots
DJ's Fisher Price Chatter
Rob-Bot's My 1St Life Size Robot Build
Woohoo! Now that's exciting stuff
. I can empathize with the challenges. Every time you feel like you're close, something pushes you a bit further back haha. But the end is close, and it's going to be very well received!
As I work my way to the end of this project I've been brainstorming about a completely local AI. My course covers all the usual ways to connect Alan and Mecha Morphix platforms for AI companions. But, as with most things today, you end up hitting a user limit when it comes to accessing these online AI voice/ audio commands.
Example: I have a paid plus Chat GPT account ($20 a month), and your audio/voice limits are: Up to 1 hour with instant or medium/high intelligence voice models, and up to 2 hours with mini voice models per 24 hours. Say Pi, (I've demoed this on my YouTube channel ) limit is even worse at 1 hour a month! That's not an AI companion, that's a closed toll gate.
I've been working on a complete local AI solution. It's open source, will work with low latency (STT /TTS) and provide you with a fairly intelligent AI with history and memory. No paywalls to worry about, and 24/7 access. It would be exciting to have on an autonomous robot roaming around and having convos with people on the street...again without the need for an internet connection/paywall limits. Since Alan can't move past the desk he is sitting on, there's no fear of him wandering into a farm field to explore a cow. Although that would be fun to see!
You'll be able to sleep at night knowing your information is local, personal, private and safe (not feeding AI companies training data).
This will likely be a standalone course coming out after the launch.
Holy cow! This would be awesome.
Oh I like this idea very much
Frxtst I developed what you are looking to do in python running on a Jetson Orin Nano and it works great. I am using openwakeword, Olama with local LLM, whisper, Vosak and piper. I also support vision with it. It also uses a vector database RAG for its memory. The response time from the voice input is almost immediate. I play the response through the TTS as it is being generated which makes it very fluid. I use a respeak mic for audio input and output as it can detect the audio input direction also.
Once you get started it is hard to stop adding features.
I call it the robot brain. The other portion in the robot is a PC that is also running Ubuntu interfaced by network cable from the Jetson to the PC. The PC code is also all python based. System can dynamically create the code for any Arduino or sensor just by telling it what you want the hardware to do and what is connected to it. After it creates the code it installs it on the Arduino making it ready for use. It is really cool to see it work. Just know if you can dream it you can make it.
I and a good friend used AI to help build this for our bots. It also has a complete bipedal walking balance algorithm built into it.
Oh _______
@smiller29 , That is VERY similar to my approach. Except I'm using Kokoro. Mic/VAD detects speech Python controller interrupts TTS STT listens LLM responds TTS resumes with new reply. Does Piper do similar? I find that natural speech illusion is broken if you can't interrupt the robot i.e barge in.
Rodney, I would love to do that but this is not the place for it. This is here to support ARC users not my software efforts.
But if you get a subscription to Claude AI and install it on your computer. You may be surprised at how fast you can develop code. I have used many different AI tools together to do some of the stuff I have made. I am also a developer myself for over 30 years. But that does not mean you can’t do it. Start off basic like using Claude, Olama with a 1-2b model and Python.
Tell Claude you want to make a AI assistant using ollama. Fast Whisper and piper. This will give you STT for voice input and a TTS for voice output.
Then add wake word support. Then tell it to make a vector RAG memory. Then just keep on adding stuff. You will be surprised how much fun it can be to tell your computer to make something and watch it write code that may work perfect or may need to be tinked with until you are happy. The more you do the smarter the AI will get.
Just give it a try and have some fun.