OpenAI pays $300M for the team behind Apple's Portrait Mode

The team that built Portrait Mode on the iPhone has just sold for $300 million. Ziv Attar and Tom Bishop left Apple to found Glass Imaging, OpenAI bought it, and the Wall Street Journal reported the price, a number that has been travelling since, including on Threads.

At a glance

• Glass Imaging was founded in 2019 in Los Altos, California, according to TechCrunch, and had raised roughly $30 million before OpenAI bought the company outright for $300 million.
• The technology uses AI to get around the physical size limits of a phone camera; its neural networks, according to TechCrunch, learn the quirks of one specific camera system.
• OpenAI invested earlier this year in Opal, which makes premium webcams, and the WSJ report on the Glass Imaging deal names no device that the camera work is headed into.

If you missed Glass Imaging on the way up: TechCrunch puts its founding at 2019, in Los Altos, California. The company's own site describes an initial seed in 2021 led by LDV Capital with GroundUP Ventures, then a $9.3 million extended seed led by GV, with Future Ventures, Abstract Ventures and LDV. In October 2024, per the same site, it opened an engineering office in Tampere, Finland.

The founders ran the team behind Apple's Portrait Mode

Ziv Attar and Tom Bishop are former Apple engineers, and before Glass Imaging they led the team that built Portrait Mode, the iPhone feature that separates a subject from the background and blurs what is behind it. A phone lens cannot do optically what a large camera does, so the software does it instead.

TechCrunch puts the money raised before the sale at about $30 million. The Wall Street Journal describes the technology as using artificial intelligence to produce image quality on par with a DSLR camera. Glass Imaging's own site states the mission this way:

to revolutionize the quality of compact camera systems on smartphones, closing the gap with professional DSLR and mirrorless cameras

Why does the AI have to learn one specific camera?

Because no two camera modules are wrong in the same way. Every lens softens the corners a little differently, every sensor has its own noise. According to TechCrunch, Glass Imaging's neural networks learn about individual camera systems rather than treating every photo from everywhere the same. Think of a tailor who has measured one person, instead of a rack of standard sizes.

What the model replaces is the ISP, the image signal processor: the chain of steps that turns raw sensor readings into a finished photo. Glass Imaging calls its version GlassAI, a neural ISP, and its site says the software has been independently tested and validated for enhancing resolution and image clarity in smartphones.

The field has an old definition to go with it. Stanford's Marc Levoy, quoted by DPReview, described computational photography as producing an ordinary photograph that could not have been taken by a traditional camera.

Earlier this year OpenAI invested in Opal, a maker of premium webcams

Opal builds premium webcams, and OpenAI put money into it earlier this year. Separately, OpenAI has been working with former Apple designer Jony Ive on a family of devices that Sam Altman has described as a new generation of AI-powered computers, according to Built In.

The same reporting, cited by Built In, puts more than 200 engineers on a range of AI-powered products, among them a speaker, smart glasses and a lamp. The speaker is said to carry a built-in camera that can identify objects and understand what is happening in a room, with facial recognition for secure payments along the lines of Apple's Face ID.

Still, the reporting stops short of the part you want: WSJ names no device the Glass Imaging technology is headed into, and no terms beyond the price. Glass Imaging's site says the work is not only software, that it co-designs novel optical systems alongside GlassAI for compact camera modules, which reads as a hardware programme. Oddly, a company whose stated goal is closing the DSLR gap on smartphones now belongs to an owner that does not sell a phone.

When the speaker is due

So how big is the chance of an OpenAI device this year? On the timelines reported so far, small: per prior reporting cited by Built In, the screenless speaker is not expected until 2027, with smart glasses likely following in 2028. Nothing in the current reporting attaches Glass Imaging's cameras to either one, and no ship date, price or product name has been given for anything built on the technology.