Super AI Creativity App Run with Local GPU on Mac/Windows/Linux [Early Access]

Last year highlighted the limitations and challenges of centralized AI services, including the spike in demand and turmoil in AI company management. At HippoML, we are developing a next-generation GPU AI inference engine can be used with NVIDIA, Apple, and AMD GPUs. We have unified the AI inference engine to serve both data center GPUs and consumer applications with a single solution. While providing enterprise customers with the best data center GPU inference capabilities, we also developed a super AI app called PrivateCanvas. This app, powered by our inference engine, aims to decentralize AI and bring its benefits to a wider audience.

PrivateCanvas: Your Cross-Platform, Free & Private AI Creativity App Powered by Local GPU (Early Access Version)

Download Now: https://hippoml.com/product/privatecanvas

Join Our Discord: https://discord.com/invite/ZP8weNy5yQ

Supported Feature:

SDXL Text to Image
SDXL LCM Fast Text to Image
Long Prompt
Refine Images
Prompt LLM
Prompt Fusion
Realtime Creating
Automatic & Manual Removal of unwanted objects
Manual Editing (Inpainting)
Expand Image with Infinity Canvas (Outpainting)
4X Upscale on Any Image

We bring state-of-the-art Large Language Model (LLMs), Stable Diffusion XL (SDXL), Stable Diffusion LCM, SDXL LCM, SDXL Inpainting, Segment Anything, and various GANs into one local app: HippoML PrivateCanvas, to help you create and edit images, entirely under your control. This is the first step towards the super AI app: our long-term goal is to allow you to process speech, video, language, images, and every AI service with your control, without depending on any single centralized online service.

High-Efficiency Optimization for GenAI

All models are compiled into a lossless low-bit format. This technology elevates GPU utilization to new heights across popular models. Moreover, an advanced model context management system enables the integration of all state-of-the-art models into as little as 8GB of GPU memory, with negligible model switching time.

Today’s generative AI applications require not just speed, but also intelligent model optimization and efficient computing to deliver breakthrough experiences. HippoML’s approach future boosts the capabilities of the NVIDIA Tensor Core GPUs while maintaining a minimal footprint, helping bring the next generation of generative AI applications to market wherever a user’s processing capabilities reside.
Ian Buck, Vice President of Hyperscale and HPC at NVIDIA

Minimal Yet Powerful Engine

No Dependencies & Portable: Our inference engine stands apart from traditional systems, requiring no dependencies on Python or other GPU libraries. They are fully portable with dynamic shape support, enabling immediate use with a simple double-click. Their compact size (~10MB — 100MB) allows for seamless integration into any application or game.

Universal Hardware Compatibility: The engine features a unified runtime and API, facilitating a single codebase to adapt effortlessly to various GPUs with optimal performance, including those from NVIDIA, AMD, and Apple.

HippoML’s Prompt LLM

Say Goodbye to Prompt Incantation

We believe that creativity should be accessible to all, without the need for “reciting a spell” to an AI model. We created a process to train Llama2 with human feedback, and created one of the best LLMs to generate prompts automatically. We enabled you to edit the LLM output to make it perfect. Leveraging our LLM technology and the standard SDXL, we’ve elevated the text-to-image experience to new heights.

Get HippoML Blog’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

We demonstrate a few prompts with Prompt LLM + Standard SDXL, Standard SDXL, OpenAI DALL-E 3 and Adobe Firefly 2.

realistic photo of a victorian young woman with steampunk makeup and colorful fine feather

Universe in a fancy Franconian bottle

Transform Simple Words into Fine Art

PrivateCanvas effortlessly transforms two straightforward sentences into exquisite art. Using dual prompts allows for richer and more intricate visual storytelling compared to a singular input. By leveraging open-source models like SDXL, we’ve significantly lowered the barriers for image generation, enabling users to craft intricate artworks with ease.

Prompt 1: Crusades under armor are running with a machine gun in desert, face toward camera; Prompt 2: desert, sand dunes, sand storms in background