Hi! I’m Leon. In this post, I want to share why Richard and I started building Owl3D and the story behind our mission to make premium 3D experiences easily accessible to everyone.
For those who haven’t heard of Owl3D, it is an AI-powered 2D to 3D video converter that adds realistic 3D depth to any video or photo. Owl3D gives you the ability to easily create 3D videos from 2D sources. You can download the software here.
The Spark of Inspiration
I fell in love with 3D a few years back in 2019 when I discovered a company called Pops (they unfortunately went out of business due to Covid) that produced lenticular prints for photos.
The 3D effect was amazing, but what truly captivated me was experiencing my personal memories in a way I had never imagined. The stereoscopic 3D added so much realism and liveliness to the photos, which made me want to see more of them. That was what brought me down the 3D rabbit hole. 🕳️🕳️🕳️
Identifying the Problems
The deeper I got into the 3D field, the more I found there were two big problems:
1. There is very little enjoyable 3D content out there besides a few 3D movies.
2. Watching 3D content isn’t as easy as it should be. We need special equipment like a 3D TV with 3D glasses, a VR headset that connects to a PC, or other 3D devices.
This got me thinking: if these two problems were solved, meaning it’s very easy to create 3D content (at least as easy as it is to create 2D content) and it’s painless to watch 3D content (like using a very lightweight headset or, even better, a glasses-free 3D screen), then theoretically we could expect 3D media to thrive and become much more popular.
The Beginning of Owl3D
This idea excited me so much. Combining it with my skills (I was a software engineer at Apple), I immediately started working on solving the first problem: how to make high-quality 3D media creation so easy that everyone can do it. I started by seeing if there was any existing solution that could do what I wanted. I played around with different VR video converters and 2D to 3D converters, but none of them met the quality standard I was looking for. I believe that 3D quality above a certain threshold is crucial because otherwise, 3D will become a gimmick.
After a few months of experimentation and many all-day-coding with my cat, I found that using depth maps generated by deep learning models to create 3D was very promising.
However, making a product out of this idea isn’t a one-person job. I needed someone with strong technical expertise who shared the same passion as I did to help me. A name came to mind right away: Richard, a software engineer at Meta and a long-term friend of mine. He was the ideal person. Thanks to his passion and courage, it didn’t take long before he agreed to leave Meta and join me full-time. This was a critical moment for Owl3D. Without Richard’s expertise, the product could never have reached its current state.
Our Vision
The long-term vision at Owl3D is to democratize 3D media, making it easily available to everyone. We believe that 3D media is a much better way to express our creativity, our emotions, and a better way to tell stories. To achieve this vision, we need to simplify 3D media creation and reduce the friction in watching 3D content.
Our first attempt at this vision was the Owl3D VR app, a cloud-based 2D to VR converter. The idea was to let people upload photos and videos to the cloud, which would handle the 3D conversion, allowing people to enjoy the converted 3D content seamlessly on their Quest headsets.
Early Success and Challenges
We spent a few months implementing this idea and released this cloud-based 2D to 3D conversion service along with the Owl3D VR app. People loved it! The magic feeling of 3D lets many people experience their favorite content in 3D for the first time! In addition, the fact that our 2d to 3d converter AI is on the cloud makes it very easy for people to get started. Given the early success with 3D photos, we soon wanted to start supporting 3D video conversion.
However, we realized a big problem (which we should’ve probably realized sooner): GPU servers were still very expensive (great for Nvidia, but not so great for us). A regular 2-hour movie has around 170,000+ frames, and for each frame, we need to figure out its 3D information. This would take a GPU server more than 12 hours to convert a movie to VR.
Given the heavy compute requirement, there was no way for us to offer a VR video conversion service in the cloud at a reasonable price for most people. We tried a few more options, hoping to find other ways to convert video to VR, but none of them made much sense economically.
Time to change
It was a tough time for us. We had to admit that even though we had put so much effort into the Owl3D cloud/VR app (at that time it had been 1.5 years into product development), it was still too early to build a video to VR converter in the cloud, and Owl3D wouldn’t be complete without video support.
We took some time to process all the feelings and confusions and decided that it was time to move on from the cloud + VR solution (for now).
