We tried to build a Robot that plays the Trumpet and Happily Failed

Sometimes Failure is The Best Option

Uri Shaked
6 min readOct 3, 2018

Every year, a group of nerds gather together for a weekend and build useless but cool things. This is Geekcon, a hackathon for makers whose sole purpose is to have fun and learn new things. It reflects well in their slogan — “If your project hasn’t failed, it means that you didn’t aim high enough”.

Last year, we tried to build the Chrome T-Rex game in real-life and succeeded: (well, sort of — it could jump and cacti would move, but no scoring and you wouldn’t die if you hit a cactus).

So this time, we learned the lesson and decided to aim higher: build a robot that plays the trumpet.

🤖 + 🎺 = ❓

Robot that plays the trumpet? Why?

I’m really into Latin music and I even bought a trumpet and tried to play it, but damn — it was hard. Getting some sound out of trumpet isn’t too hard. Getting the right sound every single time is very hard, and unlike the piano or the guitar, brass instruments are not forgiving, every single mistakes can sounds like a horrible squeak and destroy the entire piece.

I could probably spent countless hours practicing and perfecting my skill, but after spending a considerable amount of time practicing (with the supported of Ariella Eliassaf, my life partner, who had some prior musical training), I realized I was still very far from a satisfactory level.

I was practicing… and practicing… and practicing… 🎺

Then it occurred to me —what if I spent my time building the robot that would play the trumpet for me? Will the robot be able to achieve a better result in shorter time?

No one has ever done this before…

Well, almost. I googled really hard to find some prior work, and found some videos of a robot that Toyota built a few years back, but no information about how it was built or how it works, and I couldn’t even be sure whether it actually plays the instrument for real or just pretends to.

The only useful piece of information I could find was this article from 1999 (pretty amazing that it’s still on the web nearly 20 years after), which mentioned latex tubes filled with water, and some article measuring air pressure and flow while playing brass instruments. I mean, internet, is this the best you can do?

Setting up

At this stage, we decided the project was Geekcon worthy. So Ariella and I teamed up with Avi Aminov and started experimenting with latex, water, syringes, and air pumps.

We met a week before GeekCon for planning, and created some complex setup that sounded like a broken robotic didgeridoo:

You can hear the air pump humming in the background…

This preparation helped us to set the expectations even lower — eventually, we did manage to get some sound that remotely resembled a trumpet, but it had was very choppy, probably because of the inconsistent flow of the air pump. So when we arrived Geekcon the next week, we hoped that by the end of the Hackathon we will at least have some sound.

Approaching The D-Day

Avi came up with a brilliant idea for fixing the choppiness of the sound: what if we used a large plastic container, that would serve as an air buffer — similar to a capacitor in electronic circuit, to help regulate the air flow?

Avi and the trumpet, best friends ever!

On our way to Geekcon, we stopped at a local grocery store and got a plastic jar for pasta storage. As soon as we arrived the hackathon, Avi hooked it up to some lips he improvised from water-filled latex gloves, drilled a small hole in the jar, and hooked the air pump output to it, and pressed the trumpet against the lips. After a few minutes of tinkering with the position of the lips and the pressure applied by his fingers — there it was: a pure trumpet sound!

Avi, your pasta-jar trick worked!

We were really pumped up! (pun intended)

It was only one hour into the hackathon, and we already got some decent sound out of the trumpet… Goal achieved! Can we go home now?

Well, not really — getting sound out of the trumpet required a lot of manual tinkering every time, we had no control over the produced tones, and we also wanted to create fingers for our Robot so it could press the buttons. Ariella started prototyping with some 3D printed model we downloaded from Thingiverse:

Ariella spent countless hours trying to come up with a setup that would allow us to control three fingers using servo motors, while Avi and I tried to perfect the artificial lips and fix them to the trumpet’s mouthpiece.

As sunset approached, we took a break to enjoy the view!
Two fingers…
Three fingers!

As you can see above, as the number of fingers evolved, we experimented with new materials — starting with a 3D-printed setup and eventually moving to PVC Foam with wooden fingers.

Putting everything together — fingers, trumpet and some Bamba

Eventually, we had 3 fingers we could control, and while we couldn’t figure how to reliably get reproducible results from the mouthpiece, we decided to call it a day and prepare the video for display. But then, air started leaking from somewhere… and one of the servo motors stopped working.

Avi quickly fixed the leak by applying some beer (beer and coffee are provided for free, 24/7 during Geekcon), and we shoot a last video featuring two working fingers and producing world-class trumpet sounds:

So eventually, we had something to present, not what we aimed for — but definitely much more than what we initially hoped for!

A little rusty, but there it was!

Just as we finished shooting this video, a funny smell came out of the project, and — it was gone!

Apparently, another servo motor has failed and caused a catastrophic short circuit. 🔥 Perhaps using the beer was not such a good idea after all… 🍻

You only had one job! Check out the burnt chip on the bottom

So, one moment we had it working, the next moment we had nothing to show.

Failure sometimes feels Good!

Last year, we worked really hard to finish the T-Rex game before the hackathon ended, so we could display it in the fair during the closing event.

This time, we knew we were probably not going to make it, and as we were more easy-going about the project, we could help other people with their projects, check out on their project and enjoy the atmosphere. When you fail and you don’t have any project to present, you can fully enjoy other people’s projects:

They were building a string-less guitar
Some kind of a hyrbid between a violin and a trombone? by Guy Sheffer
Shooting rings of smoke
A robot that randomly walks around and teases people (yeah, it really burps at the end of the video)

and we also had some time to learn how to play the Accordion (unfortunately no photos) and even play tricks with fire (under adult supervision, of course):

Oded Sharon showing us the magic in fire! ✨

It is your turn now!

Summarizing, we had a much richer experience this year. We aimed high, we failed, we learned a lot through the preparation and the process, and we were totally cool about the failure.

Building useless things is not only fun, but also teaches you some important life lessons. Simone Giertz really nailed this in her recent TED talk — If you enjoyed reading this post, go watch her talk now.

What kind of useless/very-ambitious-ought-to-fail things have you been making recently? Or what would you like to? Please write a comment and share them with me. 🤓

Goodbye Geekcon! See you next year!

This is my 3rd post in my personal Postober challenge — writing something new every single day throughout October.

I will tweet whenever I publish a new post, promise! ✍



Uri Shaked

Google Developer Expert for Web Technologies, Maker and Public Speaker