Body Labs

Body Labs
Information
Industry	3D body modeling, computer vision, artificial intelligence
Founded	2013
Founder	Bill O'Farrell, Eric Rachlin, Alex Weiss, Michael J. Black
Headquarters	Manhattan, New York City, United States
Notable Personnel	Bill O'Farrell (CEO), Eric Rachlin (CTO), Alex Weiss (VP Software Engineering), Michael J. Black (science advisor)
Products	BodyHub, BodyKit, Body Labs Blue, Body Labs Red, Mosh, SOMA
Parent	Amazon

Body Labs was a New York City company that built statistical 3D models of the human body and software for capturing body shape and motion from scans, measurements, photos and video. It was founded in 2013 and was headquartered in Manhattan. Its technology came from research at Brown University and the Max Planck Institute for Intelligent Systems, where co-founder Michael J. Black led the Perceiving Systems department.

The company initially focused on customized apparel and sizing, and also marketed its body data for health and fitness tracking and for avatars in games, virtual reality and augmented reality. Its SOMA platform, launched in May 2017, estimated 3D body shape and motion from ordinary photos and video without motion-capture markers. Amazon acquired Body Labs in October 2017 without disclosing terms; TechCrunch estimated the price at US$50 million to US$70 million, citing unnamed sources.

Reviewed 21 October 2026. Founding, founders, funding rounds, patents, products, Mosh/SOMA technology, Amazon acquisition reporting and SMPL details checked against the cited press, archived Body Labs pages, Max Planck pages and Crossref. About review dates.

History

Research origins and founding

The company's body models were built on what it described in 2014 as "nearly a decade of research" at Brown University and the Max Planck Institute for Intelligent Systems, based on thousands of body scans. Its 2017 team page listed four co-founders: chief executive Bill O'Farrell, whose earlier start-ups included SpeechWorks (acquired by Nuance), CoSA (acquired by Adobe) and OpenAir (acquired by NetSuite); chief technology officer Eric Rachlin and vice president of software engineering Alex Weiss, both listed with research scientist positions in the Perceiving Systems department of the Max Planck Institute for Intelligent Systems and with degrees from Brown University; and Michael J. Black, the department's director, as co-founder and science advisor. The Max Planck Society's biography of Black also states that he co-founded Body Labs Inc. in 2013 and that it was acquired by Amazon in 2017. The same 2017 team page listed Matthew Loper and Javier Romero, two of the authors of the SMPL body model paper, among the company's researchers.

Funding

On 13 November 2014 Body Labs announced a US$2.2 million seed round led by FirstMark Capital, with New York Angels and existing investors taking part. Its first application at the time was BodyHub.com, a web application for building a personal avatar from laser scans, typed measurements or a standard body model.

On 3 November 2015 the company announced a Series A round led by Intel Capital, with FirstMark Capital, Max-Planck-Innovation GmbH, Osage University Partners and Catalus Capital also investing. TechCrunch, VentureBeat and UploadVR reported the round as US$8 million; later coverage by 3D Printing Industry gave the Series A as US$11 million. VentureBeat wrote that the company held a collection of 2,300 body scans that it wanted to open to developers, and that to do so its BodyKit API needed deeper integration with hardware such as Intel RealSense, so that app developers could upload body scans directly from 3D sensors. TechCrunch said Body Labs had until then worked as a consultant to brands and organizations; it was working with the U.S. Army on Kevlar vests that fit women, and with shoe manufacturers (orthotics) and sports apparel companies (protective gear for professional football players). The article said the money would go toward its own tools and products.

In March 2016 Body Labs announced exclusive rights to two Brown University patent documents (U.S. Patent 9,189,886 B2 and publication 2013/0249908 A1) covering anatomically accurate avatars generated from a statistical model of human shape, pose and motion, along with licenses from Max-Planck-Innovation that extended its models to hands, feet, faces and heads and added soft-tissue deformation from 4D scan data.

Acquisition by Amazon

TechCrunch reported the acquisition by Amazon on 3 October 2017. Amazon declined to comment, and Body Labs confirmed the deal with a note on its website. The note read: "We've been acquired by Amazon. We look forward to innovating on behalf of customers," and referred inquiries to Amazon's press office. TechCrunch put the price at US$50 million to US$70 million, noting that one source suggested more than US$100 million while a second placed it closer to US$70 million. It also described the deal as possibly an acqui-hire and speculated about uses in Amazon's fashion business. AlleyWatch reported total funding of US$13.2 million over two rounds.

Products

Product	Date	Description
BodyHub.com	2014	Web application for creating and customizing a personal 3D avatar from laser scans, measurements or a standard body model
BodyKit	Beta, March 2015	Set of APIs and embeddable components (ShapeX, Instant API, BodyHub API) to create, pose, measure, animate and compare 3D body models
Body Labs Blue	2016	Embeddable web interface and API for online clothing retailers that predicts body shape and a full measurement set from basic inputs and refined measurements
Body Labs Red	Reported 2017	API that applied AI to 3D scan data from high-end laser scanners or consumer depth sensors
Mosh	15 February 2017	iPhone photo app that detected a person's pose in a photo and added 3D effects, animated outfits and environments that reacted to the body
SOMA	May 2017	"Human-aware" AI platform predicting 3D body shape and motion from photos and video
SOMA Shape API	2017	API that the company said could predict and measure a customer's 3D shape from a single image

BodyKit and the developer APIs

BodyKit entered beta on 3 March 2015. Developers could upload a 3D body scan or a set of measurements and get back a model that could be viewed, posed and animated with soft-tissue deformation. The Next Web reported that the toolkit extracted more than 50 standard body measurements, calculated volume and surface area, applied motion-capture sequences, compared bodies over time with heat maps, and exported posed models in several formats for 3D CAD software. Its listed application areas included fitness, fashion, ergonomics, video games, animation and VR. When the Series A was announced, UploadVR described a Body Snap service that scanned a person with Microsoft Kinect.

In October 2016 Body Labs partnered with 3dMD, a maker of 3D and 4D body scanners. 3dMD scanners captured the body, and Body Labs software turned the scans into avatars with color data that could be reposed, measured, animated and brought into CAD tools, animation software, and game or virtual reality engines.

Mosh and SOMA

According to Body Labs, Mosh combined deep learning with the company's generative body model, SMPL (Skinned Multi-Person Linear Model). The system predicted major joints, toes and facial features, estimated 3D landmark locations, and fitted SMPL's 3D pose and shape parameters to them. The convolutional neural network ran on a backend server, which was needed only for pose estimation, while the 3D graphics were rendered on the iPhone. The company called Mosh a first step toward mobile apps that "augment the body" for games, communication and shopping. 3D Printing Industry described the app's effects as augmented reality features.

SOMA, announced on 31 May 2017, applied the same approach to photos and video. 3DPrint.com reported that its neural networks fitted the SMPL mesh to video, scan and photo input, and that the company aimed it at smart homes, personalized shopping, autonomous vehicles and gaming, including pedestrian action prediction and controller-free gesture commands. Body Labs' own SOMA page said the system captured motion without markers and could transfer actions from user videos "into interactive games or virtual environments".

XR work

Body Labs' relevance to VR and AR came through its body models, developer APIs and the Mosh app. 3DPrint.com's report on the 2016 patent and license deal listed computer animation, gaming, VR and AR among the workflows the new technologies would serve, together with health monitoring and fitness tracking, and said the avatars could personalize VR or AR experiences. UploadVR suggested in 2015 that scanned body models could support mixed reality changing rooms and VR health or shopping apps, and the 3dMD partnership let scanned avatars be brought into game and VR engines.

For gaming, 3DPrint.com reported that SOMA could detect motion and shape from user video and transfer a player's movements into a VR environment or interactive game, give each captured motion a skeleton for gaming and animation pipelines, run real-time AR or VR experiences, and fit MMO characters, sports players and in-game avatars to each player's body shape.

The SMPL model used in Mosh and SOMA is a skinned, vertex-based body model trained on a large number of aligned 3D meshes of different people in different poses. Its authors describe it as compatible with existing rendering engines because it is based on blend skinning. The Max Planck Institute's SMPL project page offers the model for research purposes after registration, with code for using it in Python, Autodesk Maya and Unity, a Blender add-on and an Unreal Engine plugin. Its model license covers non-commercial scientific research, and the license page states that commercial licensing is available through Meshcapade.

See also

• Avatar
• 3D scanning
• Body tracking
• Motion capture