OpenAI open sources Point-E, a machine learning system that generates and displays a 3D object from a text prompt in one to two minutes on an Nvidia V100 GPU
TechCrunchKyle Wiggers
Context & Ripple Effects
Point-E extends the text-to-asset line OpenAI started with image generation: after releasing DALL-E 2 to researchers in preview, the lab is now publishing a text-to-3D system outright under an open-source license rather than gating access. The one-to-two-minute generation time on a single Nvidia V100 GPU signals the compute bar for usable 3D synthesis has dropped to hardware many developers already own.
The release also fits OpenAI's split distribution pattern visible across its history — from GPT-3's research impact through DALL-E 3 being folded into paid ChatGPT tiers — where select models are opened while flagship capabilities are monetized through subscription products.
First-order effects
Developers and researchers can now run text-to-3D generation locally on existing Nvidia V100-class GPUs without waiting for API access, making 3D asset prototyping a self-serve capability.
Second-order effects
Nvidia gains another workload that drives demand for its installed GPU base, reinforcing the same hardware pull seen when Nvidia Research built Eureka on GPT-4 for robot skill training — more generative modalities, more GPU cycles.
Third-order effects
If the open-release-plus-monetized-flagship pattern holds, 3D generation follows the same path as imaging: community-built tooling around open weights at the bottom, while polished multimodal creation converges into paid assistant products like ChatGPT.
The trend:Generative AI is expanding from text and images into 3D content creation, with labs alternating between open-weight releases that seed developer ecosystems and closed products that monetize the polished experience.
OpenAI just dropped a prototype of 3D DALLE (called “Point-E") 👀. It isn't as good as Google's DreamFusion, but blazing fast! Like ~600x faster to generate 😮. 2D DALLE has already turned the creative world upside down. How will 3D DALLE disrupt games, VR, metaverse, ...? 🤯 https:…
OpenAI released Point-E, a text-to-3D (point clouds) demo 🔥 You can check out an open-source demo for it at https://hf.co/... The demo uses the lower-quality but much faster version of the model. https://twitter.com/...
Point·E: A System for Generating 3D Point Clouds from Complex Prompts abs: https://arxiv.org/... github: https://github.com/... https://twitter.com/...
Bonus: it's fully open-source! Both the code and all model weights (for 40M, 300M, 1B parameters) are available at https://github.com/... Arxiv paper: https://arxiv.org/... 2023 will be a year of powerful 3D generative models, and 2D will simply be their planar projection. 🤯 http…
Shipping every week! We just open-sourced Point·E - a model that outputs 3D point cloud models. Does in 1-2 minutes what would previously take hours! https://github.com/...
OpenAI just released a 3D version of DALL-E called Point-E github: https://github.com/... arXiv: https://arxiv.org/... I'm not an expert in the area, but I've seen a lot of work on 3D generation over the last few months. Anyone know if this stands out in any way? https://twitter.…