David Holz has put Midjourney's full-body ultrasound scanner on camera again, and the new look makes his strangest bet more tangible without answering the question that now defines it: what has Midjourney actually proved?
The Verge reported Friday that Midjourney released a behind-the-scenes video of the dunk-tank scanner, a machine the image-model company wants to place in spas and eventually tie to frequent, low-cost, radiation-free body imaging. The video is hosted by Marcin Plaza, a tech YouTuber who is also an engineer at Midjourney, and The Verge says it runs nearly 20 minutes.
[Midjourney Medical behind-the-scenes video](https://www.youtube.com/watch?v=4nzzpUKhj1M&ref=runtimewire)
[https://youtu.be/4nzzpUKhj1M](https://youtu.be/4nzzpUKhj1M)
For Holz, the scanner is a founder-led expansion at a company built to resist the usual venture timetable. Midjourney describes itself as a community-funded research lab of 60 people and says it is lean, self-funded and distributed. Its homepage now lists medical alongside image and video models and other software and hardware projects under themes including imagination, coordination, reflection, beauty and human flourishing. That framing matters because Midjourney Medical is being presented less like a product-line extension and more like Holz using the cash engine of creative AI to fund a different kind of laboratory.
Holz made that explicit in the video. According to The Verge, he said Midjourney's lack of investors gives him freedom to pursue the scanner. "No one can tell me not to do it," he said.
That sentence is the appeal of the whole project. It is also the governance problem. Midjourney can chase an unfashionable hardware-and-health thesis without a board forcing focus, a lead investor demanding milestones or a public market asking how a spa scanner fits beside generative image subscriptions. The same structure leaves the burden of evidence almost entirely on Holz and the operators around him.
RuntimeWire reported in June that Midjourney Medical turns Holz's company into a body-data company if the spa model works. We also covered the initial scanner reveal, where the same tension was already visible: Midjourney was selling the romance of full-body imaging while keeping diagnostic use on the far side of FDA clearance. The new video adds more metal, water and bench work. It does not add the validation data that would let the company speak about medical outcomes with the confidence of a medical device maker.
The hardware is getting easier to see
The most useful part of the video is that Midjourney Medical no longer exists only as concept art and event-stage language. Plaza shows the scanner as an assembly of ultrasound hardware, off-the-shelf computers and Raspberry Pis connected around a water-based scanning system. The Verge says he describes the build in plain engineering terms, including as a group of ultrasound probes rebuilt around what amounts to a hot tub with a lift.
That candor helps. Hardware founders often hide ugly prototypes behind polished renders until launch, and Midjourney is choosing to show some of the rough edges. The image in The Verge's story, credited to Midjourney Medical, shows a scan of an imaging phantom segmented to validate how structures separate under controlled conditions. A phantom scan is a reasonable early engineering artifact. It is not a substitute for clinical validation on people.
The unanswered questions are the same ones that matter to radiologists and operators. Midjourney has not disclosed a price for the scanner, a commercial scan duration in real use, a deployment count, spa partners, clinical trial data, peer-reviewed performance metrics or an FDA clearance number. The Verge says experts previously told it that Midjourney had shown little evidence it could overcome known ultrasound limits or produce the detailed images it has suggested at the scale and speed it is promising.
That gap is sharper because Midjourney's existing business is built on images that feel convincing. In generative AI, a beautiful image can carry the sale even when the underlying process is opaque. In medical imaging, the image is only the surface. The hard claims live in sensitivity, specificity, resolution, reproducibility, technician dependence, patient selection, artifact rates and what physicians are legally and clinically allowed to do with the output.
Wellness is the lane Midjourney chose
Midjourney is trying to enter through wellness first. The Verge says the company is emphasizing body composition rather than diagnostic use, a choice that avoids the immediate burden of marketing the scanner as a medical diagnostic device. Tom Calloway, Midjourney's head of medical, said in the video that focusing on body composition would let Midjourney "speedrun" and open once testing is complete, according to The Verge.
That is a commercially coherent path. Body composition is easier to explain in a spa than cancer screening, and consumers already pay for devices and services that promise more personal health data. It also gives Midjourney a way to collect usage, improve the customer experience and learn how bodies behave inside its scanner before it asks regulators, clinicians and insurers to take on a larger claim.
The risk is that the customer hears medicine even when the legal category says wellness. The Verge says the video still uses medical language, including discussion of what physicians might do with frequent scans over time. That is the line Midjourney will have to manage with unusual discipline. A spa visit can produce body maps. A diagnostic product has to prove what those maps can detect, for whom, under which conditions and with what downstream consequences.
The spa setting also changes the operating burden. If Midjourney wants a scanner in a San Francisco spa, the user experience must handle water, throughput, sanitation, privacy, comfort and repeat visits with the same care that a hospital handles protocols and records. If Midjourney wants physicians to eventually care about the data, it will need evidence that repeated scans are reliable enough to compare across time and clear enough to change decisions.
Self-funded freedom has a deadline
Holz's autonomy is the strongest founder color in the story. RuntimeWire previously noted his return to interfaces as Midjourney moved from models into hardware. The scanner fits that arc: it is an interface to the body, wrapped in a consumer environment, funded by a company whose public identity was built on synthetic images rather than regulated health infrastructure.
The investor-free setup lets Holz make a move that a conventional AI startup would struggle to defend. A venture-backed image-model company would have to explain why scarce engineering time is moving into ultrasound, spa real estate and FDA-adjacent workflows while the generative AI market remains expensive and competitive. Midjourney's community-funded structure gives Holz room to pursue a thesis before the business case is obvious.
That room will narrow as soon as people step into the water. Once a customer pays for a scan, Midjourney's claims become product obligations. Once doctors are invited into the story, even as potential recipients of shared data, the burden rises again. Once Midjourney suggests frequent scanning can help people live longer or avoid disease, the company is operating in a trust market where careful language matters as much as engineering velocity.
The fairest read is that Midjourney has shown a real prototype and a founder willing to spend company freedom on an unpopularly hard problem. That is worth taking seriously. The same facts make the missing evidence impossible to ignore. Midjourney Medical can be a wellness product first, a body-composition mapper first and a spa experience first. Its public ambition still points toward medicine, and medicine will ask for proof that a behind-the-scenes video cannot provide.