Midjourney’s Most Ambitious Image ‘Generation’ Yet
Midjourney, the AI Image generation leader, have announced a whole-body scanner built on proven ultasound technology and pointed at an unsolved physics scaling problem.
Every so often, an announcement from a company is so out of the norm that you check the calendar for April 1st. Midjourney, the lab by far and away the AI creative image-generation leader, has announced it is building a machine to look inside your body.
Not a model. Not an interface. An actual physical ‘whole body’ scanner that intends to map you head to toe in about the time it will take you to read this introduction.
It is called Midjourney Medical, and the device is the Midjourney Scanner. The pitch, from David Holz is the first new whole-body imaging modality in fifty years. You step down into a pool of water. Forty chips, each smaller than a postage stamp, line the walls of the tank — and here is where the story gets interesting, because those chips are not Midjourney’s.
Butterfly Network
The chips belong to Butterfly Network (NYSE: BFLY), a Burlington, Massachusetts medical device company that has spent thirteen years doing exactly one thing: replacing the piezoelectric crystals in traditional ultrasound probes with arrays of capacitive micromachined transducer elements etched onto semiconductor wafers, integrating multiple signal generation and processing transducers into a single chip.
Butterfly’s Ultrasound-on-Chip technology received its first FDA clearance in 2017. The third generation, the iQ3, cleared the FDA in January 2024. Over 145,000 clinical customers use the technology today, in emergency rooms and operating theatres, at bedsides, and in the field in traditional handheld transducer wands. Midjourney signed an exclusive licensing deal for the chip in November 2025.
Each Midjourney Scanner deploys 40 of the Butterfly chips in a circular array. Without the Butterfly array, the Scanner is simply a pool of water. The butterfly elements take turns singing and listening, millions of times a second, and then AI-driven software reassembles the echoes into a three-dimensional map of muscle, tissue, and the structures beneath.
No X-rays. No magnets. No gantry, no contrast injection, no radiographer telling you to hold your breath.
And here is the part worth pausing on, because easy cynicism misses it. There is no ionizing radiation in this thing at all. That is a huge deal. Much of the medical case against electively scanning healthy people rests on cost and on the small cancer risk you accept every time you climb into a traditional X-ray-based CT.
Midjourney is not promising a single expensive snapshot. It is promising a cheap, repeatable baseline, the kind of thing you might do yearly and compare against your own prior self, watching for changes rather than hunting one time for absolutes. A few dollars a scan, they say..(we’ll see).
This is the Star Trek sickbay scanner made real, not in 2301, but rolled out widely by 2031. Star Trek has a long history of predicting or inspiring our future technology (and issuing a few warnings about absolute trust in computers…but I digress)
The grounded medical science.
What sounds fantastical science fiction is actually grounded in technology that is quite real. Computed tomography is a well-understood, accepted, and research-grounded methodology: slices are reconstructed by computation from many measurements taken around the body, not an energy source. Its cousins make the point:
Computed X-ray Tomography (with or without contrast) is a ubiquitous technology in radiology departments worldwide.
SPECT reconstructs from gamma photons,
OCT from light (an absolutely fascinating aspect of the micro application of the tech)
Ultrasound computed tomography (USCT) is the well-established name for a technique that already exists, already helps, and already has a growing trail of clinical deployments behind it.
Revolutionizing Breast Screening
The leading system using USCT is SoftVue, developed by Delphinus Medical Technologies of Novi, Michigan, and cleared by the FDA in October 2021 as an adjunct to mammography for screening women with dense breasts. Understanding what it actually does, and what it feels like, matters here — because it is a direct conceptual ancestor of the Midjourney Scanner, and the comparison with the gold standard of standard X-ray mammography is instructive. Delphinus’s own explainer video shows you how the technology works, why it’s so diagnostically superior, and what the patient experience looks like
The patient lies face down on a padded table. There is an opening in the tabletop, and she lowers one breast into a warm-water bath beneath it, the temperature maintained at around body temperature. A soft, anatomically formed interface called the Sequr guides and steadies the breast in the water without compressing it.
Then a ring transducer, twenty-two centimeters in diameter, begins to travel along the length of the breast through the water, collecting data every two millimeters from the base to the nipple. The ring carries more than two thousand elements in a full 360-degree array, each taking turns firing sound waves and listening for the return signal. The whole breast scan takes two to four minutes. The patient does not need to move. Nothing presses on her. There is no radiation. The breast is not visible at any point during the scan, which Delphinus makes a point of noting.
What SoftVue returns is not merely a picture of structure. Its TriAD technology, which stands for Triple Acoustic Detection, captures three distinct signals simultaneously: reflection (what standard ultrasound already does), speed of sound through the tissue, and attenuation, meaning how much sound is absorbed rather than transmitted. That third category in particular, transmission data collected from the elements on the far side of the breast, is not available from any other commercially deployed breast ultrasound device. The combination gives the radiologist a tissue characterization, not just an anatomical shadow.
Just show a woman in your life the video of the procedure and compare it to the traditional (rather brutal) mammography session, and ask her which she would rather have done.
Most importantly, dense breast tissue affects around forty percent of women in the United States, and standard mammography is a cruelly inadequate tool for exactly this population. It flattens the breast between two plates and fires X-rays through it, a compression most patients endure, grateful that it is a relatively short procedure.
One that reduces breast cancer mortality by forty-one percent in women with non-dense tissue and only thirteen percent in women with dense tissue, precisely where the risk runs highest. When something looks off, the follow-up is usually a handheld ultrasound, gel, and probe, and an operator whose skill determines what gets seen. SoftVue is none of that. No plates, no radiation, no skilled operator improvising against an uncooperative body, and in the dense-breast trials, it found up to twenty percent more cancers than mammography alone, with fewer false positives and fewer unnecessary biopsies.
The system is currently installed and operating at the Barbara Ann Karmanos Cancer Institute in Detroit, East Alabama Medical Center, the University of Rochester Medical Center, McLeod Regional Medical Center in Florence, South Carolina, West Jefferson Medical Center in Louisiana, and Corewell Health, with more sites in active deployment. It is not a research prototype. It is in clinical use, generating real reads on real patients today.
So when Midjourney talks about ultrasound tomography, they are not bluffing. They are standing on a real, validated, genuinely kinder foundation, and pointing it at the whole body. It’s an ambitious and wonderful stretch in engineering and computing.
But Enough from the Enthusiast. Now the Skeptic.
If I sound grumpy, please consider buying me a coffee
Start with what they actually showed, because it is not the machine in the brochure. This is a generation-one prototype, built by a team of around nine, that has scanned perhaps a dozen people. The sixty-second scan is a target; today, the scans take about twenty minutes, bottlenecked by bandwidth, algorithms, and the plumbing that moves the data.
To their credit, Midjourney says the images shown are not AI-generated, which means what is on screen is a research result from a handful of bodies, not a product. Absolutely fair for a first reveal. But healthcare’s buyers have spent a decade learning to discount the glossy AI promise, and they did not learn it cheaply.
Things ultrasound hates
Ultrasound hates two things, and the human body is full of both. It is throttled by bone and effectively stopped by gas, and neither is a tuning problem you engineer easily around; it is the nature of the wave. Now consider what a whole-body scan is actually being asked to do. The organs you most want to see sit in the center, wrapped in the two worst possible obstacles: a cage of ribs, a vault of skull, and lungs that are full of air. This is exactly why the cleared system images the breast and nothing else. The breast is an organ that you can lower whole into water with no bone and no air in the path.
Let me make this personal for a moment because I have recent, intimate evidence.
At my annual physical not long ago, I had an echocardiogram, an ultrasound of the heart, and it took the best part of an hour. The technician worked a handheld transducer against my chest wall, inching it between my ribs a few millimeters at a time, hunting for a gap she could see the heart through. The ribs are bones, and bones are a barrier to ultrasound.
At one point, she laughed and told me my lung transplant was too good. My new lungs were so well filled with air that they kept drifting across the structures she was after and blotting them out. Her remedy was to have me breathe all the way out and hold it, dozens of times over, while she darted the probe to a new angle in the ten seconds or so before the air flooded back. A skilled human being, a handheld probe, two known enemies, one moving target, and a full hour to image a single organ.
Anyone who sat through tenth-grade physics can remember what waves do. They reflect. They meet other reflections and interfere, and that interference shows up in the image as noise. A good technician spends the whole exam fighting it, adjusting frequency, changing angle, coaxing the beam until the echo comes back clean enough to be diagnostic. That is the craft, and it is mostly improvisation against a body that will not cooperate.
Now take away the hands. The Scanner replaces the technician with an array, hundreds of thousands of elements each firing and listening, and asks software to do in arithmetic what she did by feel, not for one beam through one rib gap but for the entire reflecting, interfering, breathing whole, all at once. The breast scanner proves it is not impossible with clear soft tissue. But the further the transducer is from the actual target and has to find its path through those pesky ribs and annoying air-filled lungs, the harder the math and science become.
Who pays, and who changes the water
Start with the lane they chose, cash-pay wellness Spas, because it is the only one really open in the short term. Self-pay diagnostics already works: Prenuvo, Ezra, and SimonMed sell whole-body MRI for hundreds to thousands of dollars, no insurance, no diagnostic claim, because in this instance, the image is not the patient any more than the chart is. But they survive on high prices and low volume.
The reimbursed lane, the one where insurers pay, is locked up twice, and the second lock is the one that engineers and founders forget. Yes, you would need FDA clearance, validation, and a billing code. BUT…There is no billing code for screening healthy people with whole-body imaging, commercial plans deny it across the board, and Medicare will not cover it at any age.
Not because the imaging is unproven, but because no one has shown that scanning the asymptomatic well person makes them live longer. A flawless Scanner with every FDA blessing would still be denied for the screening it was built to do, because that is a question about evidence, not hardware.
And then there is the operations part, as yet undiscussed. The unglamorous business of running the thing at scale.
Do you drain and refill the water between every customer, and if you do, what does that do to the turnaround and the throughput that the billion-scans math depends on?
What is the calibration and quality cycle that keeps each of fifty thousand machines producing data anyone can trust, and who signs it off?
How do you handle the customer who twitches, or breathes, or drifts in the water, smearing the scan with the exact motion interference a technician spends an hour suppressing by hand?
How do you handle the substantial number of people who cannot bear to be lowered into a tank at all, the claustrophobes who already struggle inside an open MRI?
None of this is unsolvable. All of it is unsolved, and every item is a cost, a delay, or a liability the announcement quietly skipped.
Conclusion
Midjourney has done a dazzling and surprising thing: combining proven ultrasound engineering and combining it with an application of their AI modeling and processing prowess. It’s very much in the spirit of their unique corporate credo; they are putting their money where their beliefs are. I would love to know the personal motivations behind what’s driving David Holz and the team to embark on this fantastic healthcare challenge.
I genuinely want them to succeed because a kind, radiation-free baseline for everybody on earth would be a gift. Certainly, for my own regular imaging to maintain the health and surveillance of my transplanted lungs. For patients like my son who suffered years of pain and discomfort because his Pancreas is so hard to image clearly.
But wanting is not evidence. Until they can carry the image through bone and air and lung, at the price they promised, and show not just that they can see but that the seeing helps, the Scanner is a glorious instrument in search of a question the evidence will pay for.
Keep haverin’.
SM
Sources
The following sources support the factual claims made in this article, listed in order of appearance. All URLs accessed June 18, 2026.
THE MIDJOURNEY MEDICAL ANNOUNCEMENT
Midjourney. “A New Era of Midjourney.” Midjourney official blog. June 18, 2026. https://www.midjourney.com/medical/blogpost
Dent, Steve. “Midjourney, the AI image generator, is developing a full-body ultrasonic scanner.” Engadget. June 18, 2026. https://www.engadget.com/2196998/midjourney-full-body-ultrasonic-scanner/ — Source for the November 2025 Butterfly Network licensing agreement, the San Francisco spa opening date of end-2027, and the 50,000-scanner / billion-scans-a-month 2031 target.
PROTOTYPE STATUS
“[AINews] Midjourney Medical: scan your organs like you step on a scale.” Latent Space. June 18, 2026.
— Source for generation-one prototype status, team of approximately nine, around a dozen people scanned to date, current scan duration approximately twenty minutes, and Midjourney’s disclosure that images shown are not AI-generated.
DATA THROUGHPUT AND SCANNER SPECIFICATIONS
“Midjourney Medical: Full-Body Ultrasonic Scanner, 60 Seconds, SF Spa 2027.” ExplainX.ai. June 18, 2026. https://explainx.ai/blog/midjourney-medical-full-body-scanner-spa-2026 — Source for the 17 gigabytes-per-second data rate, 500,000 transducer count, and the “few dollars per scan” pricing ambition.
BUTTERFLY NETWORK — ULTRASOUND-ON-CHIP TECHNOLOGY AND CLINICAL TRACK RECORD
Butterfly Network. “Ultrasound-on-Chip™ Technology.” Butterfly Network. https://www.butterflynetwork.com/technology — Source for the CMUT element description, semiconductor chip integration, and 1–12 MHz frequency range.
Butterfly Network. “Butterfly Announces FDA Clearance of its Next-Gen Handheld Ultrasound System, Butterfly iQ3.” Butterfly Network press release. January 8, 2024. https://www.butterflynetwork.com/press-releases/fda-clearance-butterfly-iq3 — Source for iQ3 FDA clearance date, 145,000 clinical customers, and first clearance in 2017.
Napolitano, Gina. “Midjourney launches a healthcare division with ‘Ultrasonic CT,’ a whole-body ultrasound scanner built on Butterfly Network’s chip technology.” RD World Online. June 18, 2026. https://www.rdworldonline.com/ai-image-firm-midjourney-spins-up-medical-division-unveils-ultrasonic-ct/ — Source for the 40 Butterfly Ultrasound-on-Chip modules per system figure.
SOFTVUE — PROCEDURE, CLINICAL EVIDENCE AND DEPLOYMENT
Delphinus Medical Technologies. “How It Works.” Delphinus Medical Technologies. https://www.delphinusmt.com/softvue/how-it-works — Source for the patient procedure: prone table, aperture, warm water bath, Sequr interface, 22cm ring transducer, data collected every 2mm, two to three minutes per breast, no compression, no radiation.
Delphinus Medical Technologies. “Advantages.” Delphinus Medical Technologies. https://www.delphinusmt.com/softvue/advantages — Source for the TriAD Triple Acoustic Detection technology (reflection, speed of sound, attenuation simultaneously) and the procedure walkthrough video:
Delphinus Medical Technologies. “Delphinus Receives FDA Approval for its SoftVue 3D Whole Breast Ultrasound Tomography System.” Delphinus Medical Technologies press release. October 12, 2021. https://www.delphinusmt.com/news/recent-news/team-news/delphinus-receives-fda-approval-for-its-softvue-3d-whole-breast-ultrasound-tomography-system/ — FDA PMA date, adjunct-to-mammography indication, up to 20% more cancers detected.
Delphinus Medical Technologies. “Delphinus Medical Announces Installation of SoftVue Breast Imaging System at UR Medicine.” Delphinus Medical Technologies press release. January 14, 2026. https://www.delphinusmt.com/news/recent-news/delphinus-medical-announces-installation-of-softvue-breast-imaging-system-at-ur-medicine— Confirmed clinical sites, 40% dense-breast prevalence figure, 41% vs 13% mortality reduction statistic.
Radiology Business. “Dense breast imaging specialists secure $12M in funding.” Radiology Business. September 30, 2025. https://radiologybusiness.com/topics/medical-imaging/womens-imaging/breast-imaging/breast-ultrasound-investment-delphinus — Corewell Health clinical site partnership and University of Rochester corroboration.
WHOLE-BODY MRI SCREENING MARKET — PRICING AND INSURANCE STATUS
BodySpec. “Prenuvo MRI: Cost, Benefits, Risks and Comparison.” BodySpec. 2026. https://www.bodyspec.com/blog/post/prenuvo_mri_cost_benefits_risks_comparison— Prenuvo at $2,499, Ezra at $499 post-Function Health acquisition, SimonMed from $650, ACR position on total-body screening.
The Fountain. “Full Body MRI Cost in 2026: What You’ll Pay With and Without Insurance.” The Fountain. April 17, 2026. https://thefountainwpb.com/full-body-mri-cost/ — Absence of CPT billing code, universal commercial insurance denial, Medicare non-coverage.
INCIDENTAL FINDINGS RATES IN WHOLE-BODY SCREENING
Ramirez, Andrea G., et al. “How to implement a radiologist led whole-body MRI screening program.” Radiology Advances, Oxford Academic. March 10, 2026. https://academic.oup.com/radadv/article/3/2/umag014/8513624 — Cancer detection in 1–2% of asymptomatic individuals; incidental findings in up to 97% of scans.
Another problem is Midjourney's partner, Butterfly Network. This is a company that developed an interesting and validated technology for handheld ultrasound 15 years ago that never took off as it should. It peaked at its IPO via a SPAC in '21 and downhill ever since through multiple rounds of management. Can BFLY scale? They are touting this heavily and it's bumped the stock. Yet can BFLY do anything other than chips for a prototype?