#437 Meshy
About Meshy, an AI 3D model generator. The Meshy 6 model is quite accomplished at generating and manipulating models from 2D images. Text prompt generation is a little more prone to errors.
Notes
Meshy is an AI 3D model generator. 3D models can be
- generated from a prompt
- generated from an image
- uploaded (e.g. blender)
Once generated, the models can be manipulated with AI assistance:
- texturing using text prompts or reference images
- adjust triangle or quad counts, switch topology types
- automatic rigging
It is possible to experiment with Meshy on their free plan, but to do anything useful (including downloading models), a subscription is required.
I first looked at Meshy back in 2025 when their leading model was Meshy 4. I was not very impressed and did not stick around. The results achieved with their current Meshy 6 model are now quite good.
In particular, generation from recognisable IP or photos is very good, but prompt-based generation is much more prone to impossible artifacts and hallucinations.
The actual Meshy UI has some klunky vibe-coded rough edges! It is functional enough though.
Line Drawing Trial
Image generated with ChatGPT:
create a black and white outline drawing of a taiko drummer, in dramatic pose holding bachi. Only the figure, on a clear background without taiko or other elements
Not a bad image, but note the legs are not complete. This could be a common challenge for 3D model generators, given incomplete source material.
As a benchmark, let’s first test the legacy Meshy 4 model (soon to be retired). I have never been impressed by this model, and yes, the results are indeed pretty poor..
After switching to their latest Meshy 6 model, results are excellent. However, doing a reverse image search, the results are suspiciously close to Ryu from the Street Fighter and Sega Virtua Fighter Akira Yuki. I suspect strong hints in the training data!
Photo Trial
I started with this image from https://southwestfolklife.org/ken-koshio-taiko-player/
Let’s see how Meshy performs. I used default settings with their latest Meshy 6 model:
Prompt Generation
Let’s try Meshy 6 with prompt-based generation. Here’s my promot:
a taiko drummer, in dramatic pose holding bachi in a classic kamae. Only the figure, without taiko or other elements
And the result:
- hmm, not so good
- didn’t follow the instructions to only generate the figure
- hallucinated some weird garb for a taiko player
- lots of “impossibles”: foot through the taiko stand; clothing fragments that wrap around the arms; bachi fragments hanging in thin air..
