
This week, smart home device maker Wyze sent an email to its 1.3 million current device owners explaining that it didn’t plan for cloud costs needed to provide free person detection for Wyze cameras. Thus, it is hoping to offset the increased costs with a pay-what-you-want model aimed at loyal customers.
The move is a stunning departure for the company, which has been known for providing low-cost smart home devices at razor-thin margins. But, it’s clear that in trying to placate its many users who want person-detection, it is trying to walk a customer-friendly path that others in the industry have sidestepped entirely.
In short, Wyze has apparently learned a lot from Wink.

In May Wink said it would be moving from a free to a subscription-based service for its hubs, causing a bit of an uproar with its customers. After all: Many, including myself, bought Wink hubs partly because there was no fee to use them. That changes later this month as the company is adding a $5 monthly fee for customers to continue using the product.
[Interested in how companies can built connected products? Join us Tuesday, July 28 for a free virtual IoT summit.]
And now, there’s a similar situation facing Wyze customers but how the company is handling it, and the reasons for it, are very different.
In fairness to Wyze, it couldn’t plan for such costs when it began to offer the person detection features because it had partnered with XNOR.AI; a company that specializes in edge-based AI. By integrating XNOR.AI’s solution, there was no cloud component required for Wyze cameras to notify device owners when a person was spotted.
Apple’s purchase of XNOR.AI for a reported $200 million back in November completely changed all that as Wyze has since been developing its own person detection feature, which is cloud-based.
As a result, Wyze is offering a creating “pay what you want” model, at least for now. Here’s part of the email explaining this development:
“When Person Detection for 12-second event videos officially launches, you will be able to name your price. You can select $0 and use it for free. Or you can make monthly contributions in whatever amount you think it’s worth to help us cover our recurring cloud costs. We will reevaluate this method in a few months. If the model works, we may consider rolling it out to all users and maybe even extend it to other Wyze services.
We’re hoping that something really cool is born out of the mistake we made. Something totally different than what any of our competitors are doing. A method to cover the cost of premium features in a way that is truly being friends with customers, and customers being friends with us.”
The new cloud-based person detection feature will be part of Cam Plus; an optional upgrade for Wyze Cam with a more traditional subscription model that has person notifications on complete motion capture videos. This differs from what was previously a free feature as those video clips were limited to 12 seconds. Wyze hasn’t yet announced the price of Cam Plus, saying it will have more information available in a few weeks.
I don’t think this situation came about through poor planning on Wyze’s part. Perhaps it could have had a backup solution to person detection through a cloud service, yes, but that’s asking a bit much when the AI needed for this was baked into the hardware at the edge.

And that’s exactly why Wyze has to go with a cloud solution now: There are 1.3 million Wyze Camera customers that have AI hardware that can’t be used. There’s simply no easy or inexpensive way to migrate over to different hardware solutions for the devices already out there. And the economics weren’t planned for.
Using an AI chipset and models on-device has a fixed cost per device. Cloud solutions? You’re looking at how much usage a feature will get when it comes to expenses and I can’t think of a company that can accurately predict how many people will walk past the webcam facing my driveway.
Cloud-based person detection is the only reasonable option until, and if, Wyze can find a different AI edge provider and bake that chipset into future products. This may sound like common sense but I mention it because some feedback Wyze is getting on its forum about this topic suggests using a different hardware provider instead of the cloud.
Sure, finding another AI at the edge partner such as Swim.ai would mitigate the recurring cloud costs getting passed on to charitable consumers but it raises a bigger challenge for the company: How would it deploy a hardware upgrade to existing devices? The reality is, Wyze can’t easily do that without some unsightly add-on dongle or by replacing current cameras.
Regardless, Wyze’s approach is a lesson that’s likely too late for Wink to learn, since the latter company is giving current customers one of two choices. Either they can pay $5 a month to use the hardware they previously purchased, or they can skip the subscription and be the proud owner of a fairly useless smart home hub.
I suspect a lack of revenue growth for Wink due to smart speakers gaining smart hub functionality also led the company to have more limited options than Wyze has, given that the latter company is the defacto “low cost, high volume” smart home device maker.
With an approach that’s more partnership-centric between Wyze and its customers, this “pay what you want” experiment shows that the company values goodwill. It’s trying to make good on its promise that person detection would be a free feature or at least as good as is feasible given the loss of a key AI partner. And although I’d like to see IoT device makers in general properly plan their business models, I think Wyze may be onto something with this community-based approach that’s worth following if you’re a device maker in this space.
I’m not sure why the hardware solution can no longer be used. That was a business agreement when the AI chip set was Incorporated into the device and that agreement should need to be honored by any company buying it shouldn’t it?
Let me give another example. Some company buys a company that made some chips in your TV that sold them to the manufacturer of the TV. That does not prevent you from using that TV.
Is it economically feasible for a company like Wyze to field an edge HW ML accelerator, akin to Intel’s Movidous dongles?
This would be a one time purchase, confine the data to the site generating the video, could be updated via SW updates for fixes and/or improvements, all for a single price and not SaaS via cloud processing.
I think you are wrong in your assertion that the AI person detection was chip based. I owned wyze cameras for at least a year before the person detection feature was added as a freebie. after a firmware upgrade and app update it was up & running on all 5 of my existing cameras. I subsequently purchased 2 more cameras and there was indeed a production change, but it only involved the CMOS light sensor. I also don’t think the user base is taking this news particularly well. I spent a hour reading posts on the wyze community forums directed at this move and it’s astonishing to see even their staunchest die hard supporters waver even a little bit (and it’s more than a little) in their loyalty. I think wyze has had an outsized number of mis-steps and dissappointments to be critical of, but the feel of this is something different.
Lesson to be learned: don’t buy stuff that relies on the cloud.
The real question that is NOT being asked is:
How many Wyze users NEED or even CARE about “person detection”? I doubt it is over 20% or even 10%. Maybe 80% WANT it, but very few NEED it.
I’ve had several cameras for over a year and haven’t even investigated the feature.
Then again, I live in a “safe” neighborhood in a “safe” location and don’t expect to ever need the feature. I just use the cameras for informational purposes and some peace of mind.
“Free” is NEVER free. You get what you pay for. If you really need security because you EXPECT to need it, Wyze is probably NOT the solution for you.
I could NOT get a camera to connect to a Bridge from 15′ away through ONE wall/window. I have ONE bulb that won’t connect either but I’m thinking that is the socket not the Wyze system. I’ll work on those after I get my Outdoor Camera and have time to fiddle and chase the Gremlins from the system.
If I really needed those items to work, I would PAY a professional company to come in and install a system that was 100% (never can be) guaranteed to work 24/7.
Based on reviews and forums, NO security system, DIY or Professionally installed, is “perfect”. Wyze is an affordable, workable, reliable system that is worth the time and money needed to give me peace of mind.
$20 with baked in ai.
Nope. That will be a server on the Internet.
Its called bate and switch.
For all my IOT I’ve been buying tuya stuff that can be flashed with opensource firmware so that the day smart app stops working and it will,
I can reflash with another os firmware and move to an internal server. Most likely ha.
I might even preempt it and just do it while I’m moving house and kitting everything out.
With regards to the cameras,
I thought wyse cams could be flashed with open source firmware too tbh.
Why not use more open ‘edge AI’ solution like K210 chip and either use library available as open-source or develop one by themselves (instead of the cloud based one)?