Subscribe to updates

Backflip, a new venture from Markforged founders Greg Mark and David Benhaim, is off to a running start, emerging from stealth with AI-powered text-to-3D model and 3D scan-to-CAD model tools.

The star of the show here is Backflip’s “Scan-to-CAD” model. Revealed this week, the model was trained on a “synthetic 3D data set of over 100 million unique 3D geometries” to create parametric 3D models from mesh data such as point clouds. The result is a file – in either Solidworks or CAD format – which users can tweak and machine, print, or otherwise manufacture.

Should the process work as advertised, it would all but eliminate the need for complicated manual measurements of existing parts that might not have associated CAD files. That would be a huge time- and energy-saver, especially in manufacturing environments, toward which the product is positioned. Here, Backflip points to a report that shows halted production can cost organizations more than $3 million per hour.

With the text-to-3D model tool – which Backflip calls “Idea-to-Mesh” – users input a text description, then receive a printable model based on it in STL, OBJ, GLB, or PLY format.

It appears similar to HP’s generative model that was shown off last year at Formnext – and with which we had a relatively positive first date. Both Backflip and HP join an ever-growing roster of such tools looking to ride the AI wave in the 3D design space.

At first glance, both models seem to be extremely practical. We’ve yet to work with either in hand, however, so word’s out on real-world performance. Still, there’s a lot of potential here – and we’re excited to see how this all plays out.

The Scan-to-CAD tool is currently waitlisted with no set release date, although Idea-to-Model is available as a tiered subscription service, starting at $20 for the “standard” package. Check out Backflip’s website for details and demos.

Read more recent news:

License: The text of "Backflip AI Emerges from Stealth with Text- and Scan-to-3D Model Tools" by All3DP is licensed under a Creative Commons Attribution 4.0 International License.