Frequently Asked Questions

We understand the reaction. An AI image company announcing a full-body scanner inside a bathhouse does sound like the setup to a joke, and we said as much in the announcement (“a little weird and a little crazy”).

But yes: it’s real. The Gen 1 prototype exists and has scanned real volunteers under an independently approved research study. The FDA has formally responded to our regulatory submission. We’ve signed a lease in Union Square, San Francisco, and completed schematic design for the spa with our architects. We publish the details as we go: the device-development post is a good place to start.

And if you’d rather see it than read about it: our engineer Marcin Plaza filmed a raw behind-the-scenes tour of the scanner prototype: the hardware, the water, and unscripted conversations with the team building it.

Because under the hood, they’re surprisingly similar problems. The Midjourney Scanner produces terabytes of raw ultrasound data per second, and the entire product hinges on one computational question: how do you turn an inconceivable volume of messy wave data into a sharp, faithful image? That’s a giant-scale computation and algorithms problem, exactly the kind of thing our research lab exists to do.

The other honest answer: we’re a community-backed research lab with no investors, and we get to ask what’s worth building for people. We think helping billions of people actually see what’s happening inside their bodies clears that bar by a wide margin.

No. We know that’s a loaded question to ask a company like ours, which is exactly why we went out of our way on this one: every cross-section in the announcement is a real reconstruction of a real person who volunteered for our first research study. You can browse more of them in the scan gallery.

That study was reviewed and approved by WCG IRB, an independent Institutional Review Board (protocol #20260427), with signed informed consent, safety exclusions, the right to withdraw at any time, and an explicit image release for anything we publish. The full details are in the device-development post.

That’s the standard for every scan image we publish (real physics, real volunteers, independent oversight), and you can check us on it: the protocol numbers are public.

We’ve heard this one a lot (we did kind of walk into it), and it deserves a real answer.

The scanner is not a generative model. It doesn’t imagine your body; it measures it. Ultrasonic waves physically pass through you, they change shape wherever tissue density and stiffness change, and we solve the physics backwards to recover the map that must have produced exactly those waves. It’s the same family of mathematics behind CT reconstruction, not the kind behind image generation. If a structure shows up in your scan, it’s because sound waves actually bounced off it.

The sharper version of this question, the one imaging researchers ask, is whether machine learning anywhere in the pipeline could make an image look more certain than the underlying data. That’s a legitimate concern, and our rule is simple: what we show you must be supported by what we measured. Any learned components are verified and validated like every other part of the device, and we’d rather show you an honest fuzzy image than a confident invented one.

So: if you have six fingers in your scan, you had six fingers in the water.

Not at all! Midjourney Medical is a new project from the same lab, not a pivot away from everything else. We’re a research lab funded by everyday people rather than investors, which means we get to place long-term bets on what we think matters most, and we’ve been quietly working on this one for years. There are more projects behind it, too.

Yes. This is intended and designed to be a technology for all of humanity, not a product you can only find in the United States. The mission is to spread it as broadly as possible, at a price point that is approachable locally, wherever it lands.

San Francisco comes first because we have to learn to run this somewhere before we run it everywhere. The roadmap from there is more cities, then more countries.

Think of a dolphin’s echolocation, then multiply it by half a million. You stand on a platform that gently lowers you into warm water at about 5 cm per second. As you descend, you pass through a ring of roughly 500,000 tiny elements, each about the size of a grain of sand, each able to act as both a tiny speaker and a tiny microphone.

They take turns sending ultrasonic waves through your body from every angle and listening to what comes back, producing terabytes of data every second. Waves change shape whenever they cross a change in density or stiffness (water to skin to fat to muscle to bone), and a large compute cluster works backwards from all those changed waves to reconstruct a detailed 3D map of your body, down to a fraction of a millimeter.

For a look at the actual machine (probes, platform, water and all), watch Marcin’s behind-the-scenes tour.

Physics. Ultrasound travels beautifully through water and terribly through air; it’s why the sonographer squirts that cold gel on you before a regular ultrasound. To image your whole body from every direction at once, you need full acoustic contact everywhere, and the only practical way to do that is to be in water.

The water decided a lot for us, honestly. It’s warm, like a hot tub, and once your product requires people to be warm and wet, you’re most of the way to a spa whether you planned one or not. (More on that below.)

Half right. Air is the real wall: sound barely gets into an air-filled lung, so lungs are a poor target for any ultrasound, ours included. Bone is different. It reflects strongly and casts acoustic shadows near structures like ribs, which is a real limitation, but it isn’t a hard stop.

The brain is the interesting case. Our friends at Aleph have already produced remarkable Doppler imagery of the brain with the same class of ultrasound chips we use (CMUTs), working through natural windows in the skull at the temple and the base of the skull. Our system adds something: the waves from our elements can be steered directionally with beamforming, left and right and up and down, which makes brain imaging a real possibility. Honestly, the bigger challenges are ergonomic: keeping your face comfortably above the water, or the alternative of a breathing apparatus. Not something we’re promising, but not something physics rules out either.

But every imaging modality has strengths and blind spots, and MRI’s happen to be the same two things: cortical bone and air-filled spaces produce almost no MRI signal. CT sees bone beautifully but comes with an X-ray dose. Each modality earns its place, and ultrasound’s is remarkable: no ionizing radiation at all, and because ultrasound CT also measures the waves that traveled through you (not just the echoes that bounced back), it recovers quantitative tissue properties that a regular ultrasound probe can’t, like speed-of-sound and attenuation maps: actual physical measurements of your tissue, point by point.

You heard both because both are true, and we’d rather be precise about it: today, on the Gen 1 prototype, a full-body scan takes 20+ minutes. Gen 2, the version being built for the spa opening, is expected to bring that down to about 5 minutes. Gen 3, the version we’re designing for the global rollout, targets 60 seconds: you go into the water, you come out of the water, you’re done.

The speed-ups aren’t magic: they come from the machine’s internal bandwidth, more compute, and some known physics tricks we simply haven’t implemented yet. When you see “60 seconds” in our materials, that’s the destination we’re engineering toward, not a spec of the current machine. We’ll keep reporting the real numbers as they improve.

A detailed 3D map of your body composition (muscle, fat, and bone, resolved down to individual structures), uploaded to your account and viewable in the Midjourney Medical app and on the website. For a sense of what the reconstructions look like today, browse the scan gallery. Scan regularly and you get the part we think matters most: the trend. How your body is actually changing over weeks and months, at a level of detail where lifestyle changes become visible long before the mirror notices.

What you don’t get on day one, deliberately, is a diagnosis or direct access to the highest quality image reconstructions. The FDA needs to clear both due to the risks inherent in medical diagnosis or even self-diagnosis for maximum quality imagery. And to the person who asked if it will “generate” you with abs: no. It shows the muscle you actually have.

The images look a lot like MRI slices, and that resemblance is why we dreamed this up: “as powerful as MRI, as casual as a trip to the spa.” But the machine could hardly be more different. No superconducting magnet, no claustrophobic bore, no facility that costs millions per room; just sound waves, warm water, and a very large amount of computation. That’s what makes it plausible to eventually run at “whenever you want” frequency instead of “once every few years if a doctor orders it.”

Where the comparison honestly breaks down: they’re different imaging modalities, with different strengths and weaknesses. MRI has soft-tissue contrast that’s hard to beat; ultrasound sees some things MRI struggles with, and measures tissue properties MRI doesn’t, like speed of sound and attenuation. Neither one can replace the other, and we’re not trying to: we’re adding a modality to the toolbox, not retiring one.

DEXA is a genuinely good tool with decades of validation behind it, and if it serves you well, keep using it. Here’s what we’re doing differently:

• A different class of scan entirely. DEXA gives you a flat 2D projection and regional totals. The Midjourney Scanner gives you a volumetric map of your whole body, resolved down to a fraction of a millimeter: where the muscle and fat actually sit, structure by structure, not just how much. Put the two outputs side by side and they barely look like the same kind of measurement.
• The experience. One is a clinical table in a paper gown; the other is a warm bath. We think that difference decides whether people actually come back monthly.

The honest flip side: DEXA has decades of reference ranges and published validation. Ours are being built now; that’s what the verification and validation program is for, and we’ll publish as we go.

Credit where it’s due: the idea of ultrasound computed tomography has been around since the 1950s. Water-coupled USCT systems have been developed and used for breast imaging for years, and there’s a serious research literature behind the reconstruction methods. We’re standing on that work, not pretending it doesn’t exist.

What’s new is the scale and the ambition of the inverse problem. The form factor alone is a novel machine with many patented inventions inside; the complications of building it are massive, and we’re the only ones who have. On top of that: hundreds of thousands of elements, terabytes of data per second, in-house algorithm breakthroughs, and a compute cluster large enough to reconstruct all of it, at a throughput and cost designed for millions of routine scans rather than one careful study. Nobody had put those pieces together before. That’s the bet.

Honest answer: accurate enough that we’re proud to publish raw reconstructions, and not yet characterized enough for us to quote you numbers. Between now and launch, the scanner goes through formal verification and validation, the part of medical-device development where every performance claim has to be tested and documented before it ships. That runs through 2027.

This is also why our first product makes modest claims. We’d rather earn the bigger claims with evidence first, and we’ll publish results as they come, including the unflattering ones.

There is no ionizing radiation. None. Unlike X-ray or CT, ultrasound is just sound waves, the same basic physics used to image pregnancies every day for the last half-century. Nothing punctures you, nothing is injected, nothing invasive happens at all.

We’ve designed our acoustic output to stay consistent with the FDA’s guidance on diagnostic ultrasound safety limits, and the whole system runs under formal risk management: every hazard we can identify (from slippery floors to skin irritation to water quality) gets scored, engineered down, and documented before release.

A fair question, and one we take very seriously. Here’s how we keep you safe:

• The lift doesn’t lower your chin below the water line at launch.
• You can stop and exit at any moment; the scan pauses for you, not the other way around.
• The motion capture system continuously tracks the position and pose of everyone in the room. If anything looks wrong, the platform stops and raises you immediately.
• A very fast check on the ultrasound data itself independently confirms your head is above the water.
• A backup battery system takes over in a power outage, so the platform can always bring you up.
• Water temperature and quality are monitored continuously on each basin’s own loop.
• One or two staff members are dedicated to the scanner row whenever the spa is open, seconds from any room, with eyes on every system.

The platform moves at about 5 cm per second: a gentle elevator, not a ride. Most people’s complaint is that the water is too pleasant to leave.

A fair thing to want to know before you get in. Here’s what keeps the water clean:

• Everyone showers before they scan. It’s a spa rule and a scanner rule, every guest, every time.
• The water is continuously skimmed and filtered, and treated both chemically and with UV light for sanitation.
• The water is replaced entirely on a regular schedule, following well-defined sanitation codes.
• The scanner carries a set of advanced electronic sensors that monitor water quality continuously, on each basin’s own loop.

You wear swimwear, you shower first (spa rules are spa rules), and you scan alone in a private room. The scan runs fully automatically; there is no technician standing over the pool with a clipboard.

The optical sensors in the room exist for one purpose: safety. The motion capture system tracks the position and pose of people in the room so the platform can stop and raise you if anything looks off. They are not there for anything else, and the humans on site are there for hospitality and emergencies, not for watching your swim.

Right now, scanning is adults-only: our current research studies enroll adults with documented exclusion criteria: things like platform weight limits, open wounds, and active infections. At opening, we expect to set the age limit at 12 years and over. Guardians can come into the scanner room with their kids, and anyone who needs assistance can have a personal caretaker or a friend present. We do want to keep each room to as few people as needed, for your data privacy, and so the safety systems can do their job. The launch product will have its own clearly published eligibility rules shaped by what we learn between now and then.

The questions we keep getting (pregnancy, pacemakers and other implants, body-size limits) are exactly the ones that eligibility policy has to answer explicitly, and we won’t improvise those answers in an FAQ before the risk analysis is done. Where we don’t yet have evidence, the default is exclusion and caution. Expanding who can use the scanner goes through the full design-control and risk process, not a settings toggle, and we’ll publish the rules well before opening.

No, and be suspicious of any wellness device that tells you it is, because “FDA approved” is a specific term reserved for the highest-risk devices that go through premarket approval (PMA). Here is exactly where we stand instead:

In February 2026 we sent the FDA a 513(g) Request for Information describing precisely what the Midjourney Scanner is and does. The agency responded on June 4, 2026 (ref. C260013): the scanner is a Class II medical device that is exempt from 510(k) premarket review, because it’s a body-composition analyzer that doesn’t diagnose or treat anything.

“Exempt” does not mean “unregulated.” We still register our establishment with the FDA, list the device, report adverse events, follow labeling and quality-system regulations, carry unique device identifiers, and handle corrections and removals: the whole apparatus. And the exemption only holds while we stay strictly inside our stated intended use. The moment we want to claim more, we owe the FDA more. Full detail in the device-development post.

Correct on the physics, and we enjoyed this thread too. Impedance plethysmography classically means measuring the body with small electrical currents, which is not what we do at all.

What’s actually going on: the FDA classifies devices primarily by intended use, not by the physics inside the box. The regulation for impedance plethysmographs (21 CFR 870.2770, product code OMV) is where body-composition analyzers live as a category; consumer bioimpedance scales sit under the same code. Since our intended use is body-composition analysis, that’s the drawer our paperwork goes in, even though our measurement principle is ultrasound. Regulatory categories are organized by what a device is for, so it makes more sense than it first looks.

Starting with body composition instead of diagnosis isn’t us being timid. It’s a long-term strategy. A general-wellness device can reach people years sooner, and every scan teaches us what we need to responsibly pursue what comes after. Meanwhile, diagnosing disease is a genuinely higher bar: it requires climbing the regulatory ladder (510(k), De Novo, or PMA depending on the claim) with clinical evidence to match.

Will we climb it? That’s the plan: step by step, with test results submitted to the FDA along the way. But we will never let the software imply a diagnosis it hasn’t earned the right to make; that’s a risk we manage as seriously as any physical hazard.

This is the most important hard question on this page, so we want to answer it very carefully.

The launch product measures body composition; it does not screen for disease. It does not flag tumors, and it will not send you an alert that says “see a doctor.” A body-composition map, however detailed, is not a cancer screen, and treating it like one would recreate the documented harms of the whole-body-screening industry: false alarms, anxiety, and unnecessary procedures. “Minimum risk of inducing anxiety or misinforming users” is written into our risk files as a design requirement.

We knew the comparison was coming the moment we wrote the announcement, and we’d rather meet it head-on than pretend nobody said it.

The lesson of that era is that extraordinary health claims demand extraordinary transparency, and secrecy is the tell. So judge us on what you can actually check: we publish our regulatory correspondence (classification, product code, reference number, date), we named the independent IRB and protocol number behind our volunteer study, we post technical deep-dives about our own development process, and our first product deliberately makes the smallest claim we could launch with (body composition, no diagnosis) rather than the biggest one we could get away with. Also, it’s ultrasound physics that’s been imaged with for decades, scaled up, not a secret box nobody’s allowed to audit.

Skepticism about health tech is healthy. Keep it. We’ll keep publishing, so there’s always something real to check us against. Also, we aren’t taking money from investors or anyone right now!

You do. Your scans, and the body-composition data we generate from them, are yours: you can get a copy, and you can ask us to delete them at any time. They live in encrypted, access-controlled systems reachable only by authorized personnel, and anything that appears in research, demos, or blog posts is stripped of identifying information first.

Mostly no at launch, and we’d rather tell you that plainly than let you assume otherwise. HIPAA governs entities like hospitals, insurers, and clinicians who bill them. A direct-to-consumer wellness product that doesn’t diagnose disease or touch insurance generally sits outside HIPAA’s legal scope, the same as your smart scale or fitness band. This is one of the most misunderstood facts in consumer health, so we want to be the ones who say it out loud.

Here’s our position: we’re building to HIPAA’s standard now, voluntarily, while the product is small, because as our capabilities climb toward clinical use, the legal obligations arrive for real, and because the entire long-term project depends on billions of people being able to trust us with the most personal data there is.

Given who’s asking and who we are, this question deserves a careful answer rather than a slogan.

Making reconstruction better does require scan data; that’s just how the science works. Today, that data comes from research volunteers who sign an explicit, IRB-approved consent describing exactly how their scans will be used, with the standing right to change their minds about their data. That’s the bar we intend to keep: your scans are yours, and how they may be used will be stated explicitly, not buried on page 40 of a terms-of-service. Deletion means deletion. And nothing from the scanner feeds image-generation products; your anatomy is not going to show up when someone prompts for “human torso, cinematic lighting.”

We’ll publish the full data policy in plain language before the spa opens, and we want the community to mark it up before it’s final.

Not from us. We don’t bill insurance (the product doesn’t touch that system at all), and we don’t sell or share your identifiable scan data with insurers, employers, or data brokers. The only routine way your data leaves our systems is you taking a copy of it, because it’s yours.

As with everything on this page: when the formal data policy is published, this commitment will be in it, in writing, where you can hold us to it.

The honest answer is that the scanner decided for us. Whole-body ultrasound requires you to be in warm water: you change into swimwear, you shower, you soak, you dry off. Add that up and the shape of the visit is unmistakable: it’s a bath. So the design question stopped being “how do we make a medical scan tolerable?” and became “how do we make it something you look forward to?”

There’s a deeper reason too: body composition is most useful when you measure it regularly (the value is in the trend, not the snapshot), and that means the scanner has to live somewhere you genuinely want to keep coming back to. So we’re putting as much care into the spa as into the scanner: it’s part of the product, not packaging around it.

Union Square, San Francisco. We’re designing the building with our architects, Basalt, around a row of nine private scanner rooms. The design is still moving, so we’re holding the rest of the numbers (square footage, floors, lockers) until it’s locked in.

The target opening is early 2028, alongside production of the Midjourney Scanner. That’s a target, not a promise; we’ll keep publishing timeline updates as things develop.

Two kinds of things. And a caveat first: the spa design has changed since our announcement and is still in flux, so treat this as the shape of the place, not the floor plan.

The spa part. The relaxation circuit you’d hope for in a good bathhouse: saunas, hot tubs, cold plunges, steam, and showers, with somewhere quiet to sit afterwards. Whole-body ultrasound already requires you to be warm and wet, so this is less an add-on than the natural setting for the scan.

The other scanners. The Midjourney Scanner is the centerpiece, but we don’t intend it to be the only measurement in the building. We’re looking at adding other scanning modalities alongside it. None of these are finalized, but the candidates include things like an eye scan and a fast, low-field brain MRI: modalities that see things ultrasound can’t, and that fit into the same visit.

We’ll publish the real list as the design settles.

No. From the very first announcement, the design brief was a place you’d want to be even if there were no scanner: somewhere you go by yourself or with friends because it’s genuinely a great bathhouse. The scans are almost a side-effect: you barely think about them, and then one day you realize you have a library of data about your own body.

If you just want to soak, soak. We think the scanner will earn your curiosity eventually.

We haven’t announced pricing, and we won’t make one up here just to have an answer. What we can tell you is the constraint we’re designing against: the entire point of this technology is maximizing the useful information you get about your body per dollar, and a scanner that only wealthy people can afford to use regularly fails our own mission statement. Pricing will be published well before the doors open, and the waitlist hears it first.

Eventually, yes. The 24-hour part is about access: ideally, anyone who wants a scan can get one within hours, not weeks, which is a big departure from how healthcare tends to work today. Being open at 3 a.m. is part of how we get there.

We want a 3 a.m. scan to be as calm and as safe as a 3 p.m. one, so the building is designed from the first sketch to largely run itself: app check-in, a bracelet for a locker key, showers that run on a touch, scanner rooms that reset themselves, and information stations with directional speakers.

And “almost no staff” isn’t quite right. There are always enough people in the building, and one or two staff members are dedicated to the row of nine scanners whenever the spa is open, seconds from any room, standing by for questions and for anything the spa’s AI assistant can’t handle. The automation isn’t there to remove people; it’s there so the people who are present can spend their attention on hospitality instead of logistics.

Yes, there is security. Security staff are on site whenever the spa is open, at 3 a.m. as much as at 3 p.m., on top of the staff dedicated to the scanner row and the people working the rest of the building.

What we haven’t decided yet is how much of the spa stays open late. One option is that overnight hours are scanner-only, with the full spa open during the day and evening; the other is the whole building around the clock. We’ll settle that as the design and the operating plan come together, and publish the hours before opening.

The roadmap: San Francisco opens in early 2028 as the first site and teaches us how to operate this at scale. After that we start expanding to more cities alongside the Gen 3 scanner: fully custom silicon, with image quality and scan times that should be night-and-day. The ambitious end-state we’re aiming for is to be building and installing 50,000+ scanners worldwide starting in 2031: enough capacity for a billion scans a month.

Is that a wildly ambitious number? Yes, and we say so ourselves. It’s a goal we’re engineering toward in public, not a delivery date; watch us work and judge for yourself.

We did say that, and we stand behind it, as exactly what it was: a statement of vision. What David Holz, our founder, described is where decades of abundant, accessible imaging could take the world: a future where seeing inside your body is as routine as stepping on a scale, where trends get noticed early, and where lifestyle changes happen years sooner. Naming a future like that out loud is how you build toward it, and we’re not going to shrink the dream to make a headline more comfortable. Imagine artificial superintelligence is everywhere in the 2030s, but the scanner hardware needed to understand your body and health doesn’t exist. That’s not okay with us!

At the same time, we’re precise about what kind of statement it is. It’s a long-term aspiration for the technology and the field, not a claim about this scanner or this decade. The evidence on population screening is genuinely contested, many conditions have no presymptomatic imaging signature, and today’s product measures body composition. The claims we market will always be the ones we can defend with evidence, and we’ll always be clear about which is which.

Four ways, in increasing order of commitment: join the community and tell us what you think; join the spa waitlist; volunteer for upcoming clinical trials; or join the team (we’re hiring across the scanner, the software, and the spa). All of it starts from the links below.

A technology like this is intended for everyone, and we need your help to make sure it turns out that way. We are all Midjourney.