Stability AI announced 'TripoSR', an AI that can generate 3D models from images, and an easy-to-use demo was also released, so I tried it out.
On March 5, 2024, Stableity AI, which develops image generation AI 'Stable Diffusion' and other products, partnered with Tripo, which develops 3D model generation AI tool '
Introducing TripoSR: Fast 3D object generation from a single image — Stability AI Japan
https://ja.stability.ai/blog/triposr-3d-generation
????New release???? In collaboration with
@tripoAI , we have released TripoSR.
TripoSR is a new image-to-3D model that allows you to create high-quality output in less than 1 second.
Click here for details: https://t.co/3vts0AHAh6 pic.twitter.com/EUVgAnmDVi — Stability AI Japan (@StabilityAI_JP) March 5, 2024
TripoSR: Fast 3D Object Reconstruction from a Single Image
(PDF file) https://static1.squarespace.com/static/6213c340453c3f502425776e/t/65e5d4991c13be738cc1f8fe/1709560988372/TripoSR_report.pdf
TripoSR is an AI that can convert images into 3D models. A demo that makes it easy to use TripoSR has already been released, so let's try it out. First, click the link below.
TripoSR - a Hugging Face Space by stabilityai
https://huggingface.co/spaces/stabilityai/TripoSR
This time, I will create a 3D model of the image below.
Drag and drop the image into the area marked 'Input Image' and click 'Generate'.
After waiting for a while, the image with the background removed and the generated 3D model will be displayed.
The generated 3D model looks like this.
You can drag the 3D model to move it around, or roll the mouse wheel to scale it up or down. Below is an image seen from the side.
The back of the doll, which was not visible in the original image, is also generated.
This is what it looks like from below.
You can download the 3D model in
You can also view examples prepared by OpenAI from the 'Example' column at the bottom of the screen.
According to Stability AI, the recently announced TripoSR is a high-speed 3D object reconstruction model that was inspired by the paper '
TripoSR can generate detailed 3D models in a fraction of the time compared to other models, including OpenLRM . Stability AI reports that it generated 3DCG in just 0.5 seconds in a test using NVIDIA A100 . Below is a graph showing the time and quality required for inference to generate 3DCG. You can see that TripoSR, shown in blue, is superior not only in generation speed but also in quality.
Furthermore, TripoSR can operate even in environments without GPUs. Regarding the uses of TripoSR, Stability AI says, ``It can be applied to entertainment, gaming, industrial design, architecture, etc., and developers, designers, and creators are encouraged to explore TripoSR's capabilities and transform their work and industries.'' I hope they will discover the possibilities.'
The TripoSR source code and model data are available at the link below.
GitHub - VAST-AI-Research/TripoSR
stabilityai/TripoSR · Hugging Face
Related Posts:
