Backflip Launches AI for 3D Scan-to-STL

Text-to-STL service Backflip is launching a new AI model aimed at helping companies reduce factory downtime, a $50 billion annual problem in industrial manufacturing. This ambitious effort seeks to address one of the biggest challenges in 3D printing: the cost and complexity of 3D modeling. Designing a specific part requires time, money, and expertise—resources many individuals lack. As a result, the number of potential products that could be made is significantly reduced. By simplifying the process of designing the right part for the right person, the market could expand considerably.

Manufacturing facilities, especially in automotive and aerospace, rely on thousands of unique components to function. When a single part breaks, it can bring an entire production line to a halt—at a cost of up to $3 million per hour in lost output. Backflip’s AI-powered tools aim to eliminate the delays associated with manually recreating CAD models for replacement parts, helping plants get back online faster.

Trinckle has attempted something similar for fixtures, as have other companies striving to make CAD more accessible and user-friendly. Now, Backflip aims to use the so-called magic sauce of AI to tackle this problem. AI—the universal mayonnaise for investment hype and future promises—could either accelerate market growth significantly or fuel yet another wave of optimism around an immature technology.

Backflip is launching a SOLIDWORKS plugin and a web app designed to help firms create engineering parts more easily. While text-to-STL has proven effective for generating fun and creative files, producing parts with precise dimensions that can properly mate with other components has remained a challenge.

The SOLIDWORKS plugin aims to convert 3D scan data directly into native SOLIDWORKS files with a feature history, while the app transforms scans into parametric CAD models that can be easily modified. This approach eliminates the need for manual measuring and recreating parts from scratch, significantly reducing downtime. While 3D scan repair software like Geomagic has been around for some time, and simpler 3D scanning solutions like Comb exist for orthopedics, Backflip’s AI is designed for broad industrial applications, making it a potentially transformative tool.

In general, resurfacing a file is easy to do—badly. However, resurfacing point clouds and other scan data with precision, so they can be effectively used for technical parts, has been a challenge. This process typically requires human intervention to ensure accuracy and usability

“3D scanners map the surface of an object with incredible precision, quickly generating millions of data points, but they produce micro surface textures that can’t be manufactured with traditional tools. Our technology automatically converts these intricate surfaces into clean geometries designed for existing 3D CAD and manufacturing software,” said Greg Mark, Backflip CEO and Cofounder.

“These new developments are the result of Backflip’s continued R&D investments in neural representation of 3D data. We’ve paired state-of-the-art AI/ML techniques with the team’s deep background in manufacturing to create a really powerful technology foundation. Our additional focus on resource-efficiency enables us to advance rapidly, and yields 60x more efficient training, 10x faster inference, and 100x the spatial resolution of existing leading methods,” said Backflip’s Head of AI, Logan Ford.

Plugin

The plugin aims to show users how a part has been created while providing even beginners with options to easily adapt the model. It also generates a native feature history, which allows engineers to tweak the design as needed and helps newer users learn how CAD parts are structured. By breaking down the process step-by-step, Backflip is not just automating CAD but also making it easier for more people to enter the field.

The underlying model has been trained on the “world’s largest” synthetic dataset of over 100 million unique 3D geometries, though the company has yet to disclose an official release date.

For decades, converting scans into functional 3D-printable files has been a significant challenge. While progress has been made, and automated workflows exist for specific applications—including medical ones—a universal scan-to-STL workflow for all types of scans and working parts remains an incredibly difficult problem to solve.

The key challenges lie in adapting data, stitching, and resurfacing scans without introducing unwanted changes—tasks that seem nearly impossible to fully automate. If the team has truly solved these issues, it would mark a major advancement for the industry. However, it remains unclear whether the system can process mobile phone scan data or which types of scanners it supports.

A broad range of users could benefit from this technology if it works well with simple scanners. If it can reliably produce engineering parts, it will be a major advantage. However, any tedious, time-consuming failure—where a part turns out unusable after hours of processing—will quickly discourage adoption.

I’ve been using Backflip extensively and am impressed with the variety of parts it can generate. If it truly enables seamless creation of engineering and technical components, I’ll be more than impressed.