Some people are not satisfied with a webcam from their notebook or an old webcam they bought many years ago.

My brother decided he wants to try streaming. Some people can just do a thing and other people want to research all the bits and pieces beforehand. Our chat history is comparable in length to all my other chats combined, so I learned a lot of it as well. The craft is full with small details and inconveniences.

He found that there are basically no good webcams. A lot of the time an image quality is meh, and the rare presence of 4k is usually not better than 1080p, because resolution is not quality. Some camera apparently does upscaling from 1080p to 4k in its driver, indicated by poor quality and high CPU load. He also has a really good camera, Fujifilm X-T1. It can't be used as a webcam when connected to a USB port. Usually it's possible to buy an HDMI-to-USB converter, but his camera can't output a live stream into HDMI, only a recorded video.

So we had an idea — maybe it is possible to create a good webcam ourselves. Would anybody care? Well, it’s almost 2021 now, streaming is still on the rise, pandemic is going strong and video meetings are everywhere. People are willing to spend some money, as indicated by countless videos on YouTube regarding the problem.

Let’s start with defining what is a good webcam.

A good webcam, in my opinion, is a device that provides a decent sound quality, at least on par with recent MacBooks, and an image quality on par with top smartphones, such as iPhone 11/12, Huawei P40 Pro, etc. As smartphones have a good image quality, then we could probably manufacture a webcam from smartphone parts that are cheap and easily available. How hard can it be?

If you’ve read at least a couple of articles on the topic of video production, you will know that to make quality content from the technical standpoint you need three things, in that particular order:

1. Sound
2. Light
3. Video quality

Let's discuss these properties and later how to manufacture it.

Sound🔗

If you have a recent MacBook, it provides a decent sound quality with its three-mic array. I haven't researched it and don’t know much about other notebooks' audio quality. If you have a desktop PC, you have to have some external mic anyway.

If the web camera has shitty audio quality, you’ll have to buy additional gear. When I researched the topic, one of the recently released webcams (available from October 2020) with RRP of $180 was producing a sound as if a person sits in a big plastic barrel with a closed lid. Their voice was very muffled with lots of resonant rattles. That level of quality wasn’t acceptable 10 years ago, how this piece of gear could have been produced in 2020 escapes my understanding.

There are many good enough USB mics for $50+, you also can buy a great XLR microphone for $100, but then you’ll need an external sound card with phantom power supply. A decent one can easily cost you $200.

With almost any of these options you’ll have to install and configure some software that will provide sound level compression, noise suppression (because your notebook fans are spinning like hell) and won’t produce clicks in the process.

I would like my webcam to do this by itself, no rattling, minimized echo, etc.

Light🔗

A lot of what makes a picture great comes from lighting the scene in a correct way.

The simplest thing is to sit in a room so the window is in front of you and the Sun isn’t shining into a lens from behind your shoulder.

Webcam cannot provide decent lighting capabilities by itself, so I won’t spend much time here. There is a webcam by Razer with a built-in LED circle, but it’s necessarily small, always shines from a webcam direction and is mostly a gimmick.

We can rely somewhat on the good sensor here. If it provides good dynamic range, HDR capabilities and good low-light performance, then a patch of sunlight on the wall won't turn everything else black, or a poorly lit room won't turn into a gray noisy picture with a bleak silhouette.

Image quality🔗

It’s impossible to get a good lens and image sensor into a notebook lid, because lids are very thin. They’re often something like 4 mm thick all together, and there is no escaping physics. Of course, it's possible to put a webcam into a thick part below the lid, but then it'll be useful for expressing anger at our colleagues with wide-open nostrils and not much else.

Maybe some crazy engineer in the future will develop an array of small sensors with an array of tiny lenses and somehow combine all that into a good image quality, but we don't have this tech yet. Any smartphone with a good camera has that camera with an ugly protruding lens as a thickest part of the phone.

So, if you want something better than a notebook's webcam, you can get some Logitech webcam with an image quality going back to 2012. It’s still better than a laptop, but nowhere close to current smartphones.

What if you want better still? The next option is to install an app on the phone and use it as a webcam. This gives a very good image quality, but it’s very inconvenient and in some applications doesn’t work at all. For some reason it isn’t always presented in applications as an option when choosing a video source.

I don’t want to constantly attach, detach, charge my phone and tune the position of a small tripod. What’s the next option?

The next option is to just buy a mirrorless photo-camera with a big fat 4/3” or even APS-C sensor. It will cost $500+ for a new one (you can buy used for cheaper). You will also need some gear to hold the camera in place, as it won’t just rest on the top of the screen like a regular webcam. You also need an HDMI-to-USB video grabber, and you need to be sure that the camera you buy can output full resolution live onto that HDMI. Also make sure that it's possible to turn off all on-screen info like 1/100s F3.5 ISO 800, which is useful when you're shooting photos and videos through viewfinder, but not so much when you're in a video meeting. This feature is called "Clean HDMI".

Of course, you will have a lot of control over the whole stuff — set a diaphragm, get good bokeh, set ISO, set white balance, etc. Do you actually need it? Well, some of you do, and a lot of you do not. I learned all that stuff when I was 16, but now I’m pretty happy with how my phone chooses everything for me and just does great photos.

All that sounds pretty involved, isn’t it? It is a lot of time to research what you need, how to set it all up, and a lot of money too.

Manufacturing it🔗

So there is a market gap between so-so webcams for $100-200 and a full-blown setup with a mirrorless camera, an external mic and lighting panels that will cost almost a grand or two, if you're so inclined.

I thought that it is a good idea to create an "upmarket" webcam for $250-300 that will provide good audio and video quality, on par with current smartphones, actually using the same parts. As smartphones are made in tens of millions, all relevant parts should be cheap and easily obtainable.

I quickly contacted some people who could help me do the project — firmware engineer, electric engineer and a company that does mechanical design.

What features are needed for a good webcam for streamers and video meetings? Audio: good parts (mics and ADC), good mechanical design and assembly (so nothing would rattle), and some post-processing. We certainly want some beam-forming capabilities to be able to focus on the person in front of the webcam, so we will need more than one microphone, maybe up to 4. These parts are easy to obtain, there are many in stock, it's not a problem.

Even if I think that 4k is not needed, I’m pretty sure it’s really a requirement to have in an upmarket webcam, as it will be really hard to sell a pricey webcam that can only do 1080p.

Also would be very nice to have some neural network inference on the chip, so we can run some AI stuff directly on the webcam without bothering a user’s computer that can be busy with other things.

I easily found a three-piece assembly for the iPhone 11 Pro for fifty bucks, two-piece from a Huawei P40 Pro for twenty five and a one-piece from P40 Lite for ten dollars. That’s with a lens, and bought in retail. Does it mean it costs peanuts (less than $5 for a sensor) to buy in bulk to produce my own webcam? Well... Maybe. I can’t just go and buy a couple thousands of these sensors. I’m pretty sure I can buy a couple thousands of replacement parts for phones, but sensors themselves are not sold freely. It's easy to buy some "random" sensor that was around for years and most certainly has "meh" quality by today's standards, but we need a good one, with good dynamic range, HDR capabilities and low-light performance.

I wrote to several vendors asking them what are the prices and conditions, and basically their answer is either silence or “we’ll sell it if you will buy hundreds of thousands per year”.

I've found suitable System-on-Chips, and it's the same — if we want to have hardware 4k encoder, it's either "sorry, it was a demo part and we don't do it", or "we have limited support capabilities, so we can only sell if you will buy hundreds of thousands, maybe you want some SoMs from our partners?". It is actually possible to buy System-on-Modules even in very small quantities, but it's going to be pretty expensive and will be good for a prototype, but for production it will blow past our budget. System-on-Module will cost almost the same as the web camera itself.

Oh, and a lead time (from order to shipment) is systematically 16+ weeks. It means they have no stock and will place an order at semiconductor foundry, have it produced, packaged and sent to customer. Swiftly.

It means I cannot just build five hundred webcams, see how good are they selling and build a couple thousand, and so on.

There is also a plastic injection step that requires upfront investment to manufacture molds, and of course design and engineering costs. All combined it was tens of thousands of dollars. I was willing to risk it and was pretty sure that in the worst case scenario it would be just a cost of my experience building hardware product.

But if I have to commit to a hundred thousand units or more, it means I would need to pour millions of dollars (that I don’t have) to manufacture an obscene amount of devices that I don’t know how to sell.

I thought about it for a couple of days and understood that I approached the problem from the wrong angle at the beginning.

The actual problem is not how to build a good web camera. The real question is how to sell it in quantity! Is it possible for me to sell hundreds of thousands of pricey webcams per year? If I know I will sell it, I can find an investment for that.

Let's describe my situation here. I’m a co-founder at prophy.science which is a SaaS, and I work on it full-time. I can probably spend 20% of my time on something else, but I certainly don’t want to abandon it completely. I also have two small beautiful kids, and I really don’t want to move to Shenzhen to organize a big production there.

Going to market🔗

Ok, what’s the current size of the market? Luckily, there are public companies in the market, so we can go look at their reports to see if there is anything meaningful. And there is, Logitech says they have about $130M in revenue from sales of PC webcams. I think they’re the biggest player in the market by a long shot, and they sell something like a few millions of webcams a year. That’s... not much?

It means that I have to sell like 10% of the Logitech shipment size, right from the start.

Sorry, what?

It’s an established player in the market with all the distribution in place, with a very recognizable brand, and here am I, knowing nothing about how to sell hardware products, and there is seemingly no way to slither into the market and scale.

Well, maybe there is, but I don’t see it. Maybe I could pull it off by diving deep down and going full-time on the problem for years, but it seems much more risky now.

Also I should mention that this market is basically indefensible. What prevents established players from doing pretty much the same, when they will see that there is a good niche? They just go to the same suppliers, get same or better parts/prices and crush a startup using their established distribution channels.

I remember a staple phrase from a lot of interviews, something along the lines of "If I knew how hard it would be, I wouldn't do it". Here we are, knowing how hard it will be, and not doing it.

I'm just going back to my SaaS, which is much more defensible, thank you very much.

My brother actually started streaming, so you can go watch him code a snake game in ClojureScript and React.

Answering the question in title: you can't buy a good webcam because existing players seemingly don't think it's a good opportunity, and for newcomers the market is hard to enter.

We're also lucky that camera manufacturers added clean live HDMI into comparatively cheap cameras, so we don't need to spend $3000+ to set it all up.