Forum adverts like this one are shown to any user who is not logged in. Join us by filling out a tiny 3 field form and you will get your own, free, dakka user account which gives a good range of benefits to you: No adverts like this in the forums anymore. Times and dates in your local timezone. Full tracking of what you have read so you can skip to your first unread post, easily see what has changed since you last logged in, and easily see what is new at a glance. Email notifications for threads you want to watch closely. Being a part of the oldest wargaming community on the net. If you are already a member then feel free to login now.

Just curious, anyone know of or have experience with giving LLMs a few photos and perhaps 3D scans and asking them to create a file for printing?

Excuse my ignorance, but what's an LLM?

 Lathe Biosas wrote:
Excuse my ignorance, but what's an LLM?

Large Language Model, a class of artificial intelligence system. Think ChatGPT and its cousins.

I haven't looked into it myself, only photogrammetry.

Haven't tried it myself, and from what I've read about AI-created meshes, it's unlikely to give you anything usable for printing.

AI is moving so rapidly that if it isn't doing what you want now, just wait 6 months and try again. I use it at work and I am absolutely astounded by the developments in free use compared to only a year ago.

 JamesY wrote:
AI is moving so rapidly that if it isn't doing what you want now, just wait 6 months and try again. I use it at work and I am absolutely astounded by the developments in free use compared to only a year ago.

^Yes, this. AI is moving crazy fast right now. I know I'm out of date because I haven't looked into any AI 3d stuff in 6 months or so.

As for cleanup of any generated models, my immediate suggestion would Blender's remesh modifier.

I wonder if the LLMs would have an issue with IP infringement.

 Lathe Biosas wrote:
I wonder if the LLMs would have an issue with IP infringement.

haha. Typically they have none.

I keep seeing ads on Reddit for Tripo3D, which claims to be able to generate 3d models from images and text descriptions, no idea if its claims are exaggerated or not, but I imagine theres some limitations.

 Lathe Biosas wrote:
I wonder if the LLMs would have an issue with IP infringement.

Not yet, but only because the law around copyright and AI hasn't been tested yet. That will change over the next few years, and I suspect (given that Disney have now weighed in) that the end result will eventually be a lot more constraints on AI output to reduce the blatant IP infringement that is happening at the moment, particularly in image generation.


Automatically Appended Next Post:

Fictional wrote:
I keep seeing ads on Reddit for Tripo3D, which claims to be able to generate 3d models from images and text descriptions, no idea if its claims are exaggerated or not, but I imagine theres some limitations.

A quick google search turns up a bunch of 'marketing articles disguised as reviews' and no actual reviews or examples of generated meshes, so I would assume it doesn't actually work particularly well.

The problem is that LLM's are basically autocorrects on steroids. Sure enough, they will produce an unlimited amount of renders, but I seriously doubt any of them would print correctly

It would require a dedicated machine learning algorithm built from the ground up for designing 3D prints..