NVIDIA Picasso

From AI Wiki

Introduction

NVIDIA Picasso is a cloud platform engineered to create and deploy AI-generated visual applications encompassing images, videos, and 3D content. Aimed at enterprises, software developers, and service providers, NVIDIA Picasso allows users to execute inference on their models, utilize NVIDIA Edify base models to train on proprietary data, or employ pretrained models for generating content from text inputs. The service is fully tailored for GPUs and streamlines the processes of training, optimization, and inference on NVIDIA DGX Cloud.

NVIDIA Edify State-of-the-Art Generative AI Foundation Models

NVIDIA Edify comprises a set of AI-driven core models that organizations and developers can either train on their proprietary data or begin with models pretrained by NVIDIA's top partners.

Image Generation from Text

Edify features a sophisticated denoising network that can produce photorealistic 4K images based on text prompts.

Video Generation from Text

With the integration of temporal layers and a state-of-the-art video denoiser, Edify is capable of creating high-quality videos that maintain temporal consistency from text inputs.

3D Generation from Text

Edify also offers an innovative optimization framework to generate 3D objects and meshes with exceptional geometric quality, all derived from text prompts.

Enhancing Applications

NVIDIA Picasso provides a wealth of opportunities for application creators. By integrating custom generative AI models, applications can stand out, and in-house teams can offer customers top-notch generative AI tools for creative exploration. The powerful inference optimizations of NVIDIA DGX Cloud enable the delivery of interactive experiences while minimizing cloud inference expenses. Furthermore, users can incorporate all their creative assets into NVIDIA Omniverse for the creation of lifelike 3D virtual environments.

Key Features

  • Cutting-edge pretrained Edify models or custom dataset training, hosted on NVIDIA DGX Cloud for inference via APIs
  • Superior performance optimization on NVIDIA DGX Cloud for unparalleled training and inference
  • Pretrained models using licensed, enterprise-standard data to ensure data authenticity
  • Streamlined APIs for effortless deployment

Key Collaborations

Adobe

Adobe is joining forces with NVIDIA to develop AI-generated models that emphasize responsible content attribution and provenance, aiming to expedite workflows for creative professionals and marketers. These models will be integrated into Adobe Cloud's flagship products, such as Photoshop, Premiere Pro, and After Effects, as well as NVIDIA Picasso.

Getty Images

As a prominent global visual content creator and marketplace, Getty Images is collaborating with NVIDIA to offer custom-designed image and video generation models on Picasso. These models, trained on fully licensed data, can be accessed by enterprises through API calls.

Shutterstock

Shutterstock is partnering with NVIDIA to devise models for generating 3D assets using fully licensed content from Shutterstock's collection. These models facilitate the creation of high-quality 3D assets from straightforward text prompts, which can be used in a variety of 3D workflows, including game development and animation.

Client Success Stories

Runway

Runway harnesses AI to produce videos in any style, emulating specific styles based on provided images or text inputs. This flexibility empowers creators to experiment with novel approaches to filmmaking and content production.

Seyhan Lee | Cuebric

Cuebric employs generative AI to design and modify virtual productions, including the creation of environments using AI-generated models. This cost-efficient approach allows filmmakers, production studios, and artists to collaborate with CGI specialists much earlier in the post-production process.

WOMBO

WOMBO Dream is a mobile application that introduces generative AI to a broader audience