Midjourney unveils full-body ultrasound scanner and spa
Midjourney is building a full-body ultrasound scanner with Butterfly Network and plans a San Francisco spa; it aims for more than 50.
TL;DR
- 01Midjourney is building a full-body ultrasound scanner with Butterfly Network and plans a San Francisco spa; it aims for more than 50.
- 02Midjourney is building a full-body ultrasound scanner in partnership with Butterfly Network and plans to open a San Francisco spa by late 2027 to house the device.
- 03The company says a single scan takes about 60 seconds and it aims to deploy more than 50,000 scanners worldwide by 2031.
Midjourney is building a full-body ultrasound scanner in partnership with Butterfly Network and plans to open a San Francisco spa by late 2027 to house the device. The company says a single scan takes about 60 seconds and it aims to deploy more than 50,000 scanners worldwide by 2031.
What is Midjourney building?
Midjourney is developing a full-body ultrasound scanner that it will debut inside a company-run spa in San Francisco, and it wants to scale to more than 50,000 scanners running worldwide by 2031. So far, only about a dozen people have been scanned, and the firm says the spa should open by the end of 2027.
The startup describes passengers entering a shallow pool where the scan begins; the project was built in partnership with ultrasound technology company Butterfly Network. Midjourney plans a third-generation scanner in 2028 with "completely custom" silicon it says will deliver a "night-and-day" jump in image quality and speed.
How does the scanner work?
The system uses a ring of underwater sensors in a shallow water tank that emit ultrasound waves and collect returning echoes, then a compute cluster converts that data into a 3D body image in about 60 seconds. The hardware contains about half a million tiny elements, each described as the size of a grain of sand, that act as both speaker and microphone.
Those elements send sound waves and measure how the waves change as they pass through skin, fat, muscle, and bone. Midjourney says the system creates radiation-free 3D images and will start by producing "body composition maps" that do not require FDA approval; the company plans to submit test results to the agency on a rolling basis to pursue diagnostic clearance later.
Why does this matter?
The scanner promises a new route into medical imaging that Midjourney positions as faster and radiation-free compared with MRI, and the company frames large-scale early imaging as a public-health lever. CEO David Holz claims that with enough early imaging the world could avoid "30% of all deaths and 50% of all healthcare costs."
Those are sweeping assertions tied to an ambitious rollout: Midjourney says the fleet could reach a throughput capable of one billion scans per month. The near-term approach — shipping body composition outputs that do not need FDA clearance while collecting clinical data — is a clear regulatory play, but it leaves open whether the system will meet diagnostic standards when Midjourney later seeks clearance.
What to watch
Key milestones to track are concrete and dateable: the first spa opening by the end of 2027, the third-generation scanner slated for 2028, and the company’s stated target to have more than 50,000 scanners operating by 2031. Also watch for the rolling submissions of test results to the FDA and any public release of validation data beyond the roughly dozen early scans Midjourney has disclosed.
If Midjourney publishes independent image-quality metrics or peer-reviewed validation, that would move the project from curiosity toward clinical plausibility. Absent those data, the spa and the early demos will remain the primary public window into how the technology actually performs.
Written by The Brieftide · Source: The Decoder
The Brieftide Daily · 06:00
Briefs like this one, in your inbox every morning.
Continue reading
More in Multimodal AIThinkDeception: Progressive RL framework for multimodal deception
ThinkDeception on arXiv uses MLLMs, a step-by-step multimodal Chain of Thought dataset and a four-tier progressive RL trainer for.
Visual-Seeker: visual-native multimodal search surpasses rivals
Zhengbo Zhang and 12 co-authors submitted Visual-Seeker on 13 Jun 2026.
Gemma 4 12B: unified, encoder-free multimodal model for laptops
Google DeepMind’s 12B model brings encoder-free vision and native audio to laptops, runs on 16GB memory and is released under Apache 2.0.
Hugging Face Spaces agents.md: chain image to 3D splats
An agent used two Hugging Face Spaces and their agents.md files to auto-generate images, reconstruct 3D Gaussian splats.