Mesh-4 (Video=Mesh)
Artificial intelligence (AI) startup Meshy has unveiled a sci-fi-quality 3D generative model that features cleaner surfaces and improved geometric detail.
VentureBeat reported on the twenty second (local time) that Meshy has released ‘Meshy-4’, the most recent AI-based 3D modeling tool that generates high-quality 3D content from text and image input.
Mesh-4 generates 3D objects with very clean and solid surfaces with none flaws, in addition to with bumpy surfaces. It also has significantly improved geometric details, allowing for tremendous details that were difficult to implement with existing AI-generated models.
To boost the creation results, â–²the strategy of converting text right into a 3D model was separated into two stages: modeling and â–²texturing.
The modeling phase creates 4 texture-less models based on text prompts. The modeling phase creates a clean, highly detailed mesh that serves as the premise for the ultimate 3D model.
The texturing stage involves choosing the generated mesh and generating high-quality textures, making a visually consistent final product.
Specifically, by separating modeling and texturing, you have got more control over the tip result. For instance, you’ll be able to edit the mesh before generating textures, or customize multiple color variations for a similar model.
We have also added a ‘Retry’ choice to the tool that creates 3D models from images, allowing users to quickly create a brand new model if they don’t seem to be satisfied with the initial results.
The advantage of the retry feature is that if the initial results don’t prove as desired, you’ll be able to regenerate the model without spending additional credits.
More importantly, you don’t should wait for the generation to be accomplished. You may quickly check the model through a preview slideshow inside 15 seconds, so you’ll be able to evaluate whether the result meets your expectations. If you happen to don’t prefer it, you’ll be able to click the Retry button to generate a new edition. Nonetheless, the Retry feature is barely available to subscribers since it uses quite a lot of computational infrastructure.
Meanwhile, the technology to generate 3D assets based on 2D images or text prompts is a field that many firms have jumped into. It is because there’s high demand for it in areas akin to game production and e-commerce, and it may well generate revenue instantly.
This month, Meta unveiled VFusion3D, a model that generates high-quality 3D content from a single image or text description. It features training using synthetic data generated by a man-made intelligence (AI) video model to deal with the shortage of 3D training data.
Meta also introduced ‘3D Gen’, an AI tool that generates high-quality 3D assets in only one minute with text descriptions last month. Then, Stability AI released ‘Stable Fast 3D (SF3D)’ model that generates 3D videos in only 0.5 seconds earlier this month.
In March, NVIDIA introduced ‘Latte3D’, which might generate 3D objects and animal images in real time in only one second from a straightforward text prompt, and Stability AI also released an AI model capable of making 360-degree 3D rendering videos.
Reporter Park Chan cpark@aitimes.com
