Haiper To Advance Generative AI Research and Development with NVIDIA Technology

Haiper To Advance Generative AI Research and Development with NVIDIA Technology

As Haiper scales to become a European generative AI powerhouse, NVIDIA technology will aid optimisation of Haiper’s proprietary perceptual foundation model for video generation.

Haiper, the London-based AI firm building a powerful foundation model for visual content, today announced it is  advancing generative AI research and development with NVIDIA technology, Haiper will further extend its video-generative AI capabilities and facilitate longer and higher-quality video production using NVIDIA technologies including  NVIDIA NeMo platform and NVIDIA NIM inference microservices, as well as the NVIDIA TensorRT library and NVIDIA Triton Inference Server software. This will further enhance the performance of its perceptual foundation model, Haiper AI. Haiper also plans to deploy its models on NVIDIA Blackwell GPUs to supercharge inference speeds. 

This will enable Haiper’s extension of its high-quality video generation to length-agnostic video generation, and follows the platform’s recently announced $13.8M seed round, led by Octopus Ventures, and the company plans to raise a Series A round in 2024.

An all-in-one perceptual foundation model, Haiper enables everyone to easily generate high-quality video content. Through its free app and website, Haiper’s proprietary models for visual generation are available today in an intuitive workflow and UI. Haiper’s platform allows users to reimagine content visualisation, allowing advanced creative tools to be accessible to all, and empowers captivating video generation across a vast range of use cases.

Building models to understand the world’s inherent physics

Haiper will harness NVIDIA NIM and NeMo, allowing the Haiper platform to supercharge its perceptual foundation model capabilities. NVIDIA NIM microservices will enhance Haiper’s inference capabilities, while NeMo, an end-to-end platform for developing custom generative AI, will scale Haiper’s generative AI training and deployment. With this increased capacity to handle larger and more complex video generation, Haiper will be able to better deliver for its extensive, growing user base.

Haiper will further optimise the performance of its video generation models using NVIDIA TensorRT and NVIDIA Triton Inference Server to deliver low latency and high throughput for inference applications, providing unmatched efficiency and speed of generation for Haiper’s videos. Haiper also plans to deploy models using NVIDIA’s recently launched Blackwell GPU, the world’s most powerful chip enabling AI training and real-time inference for models scaling up to 10 trillion parameters, meaning Haiper can continue to build and iterate its perceptual foundation model.

Discovering further Generative AI use cases

For video-generative AI to reach the next stage, companies will need to scale models and the data behind them to better understand the inherent physics of the world. Each frame of a video carries an array of minute visual information, including light, motion, texture, and interactions between objects. Generative AI that can intuitively understand and replicate the emotional and physical elements of reality will then be able to create content that is visually stunning and true to life. Already making continued advancements to its powerful foundation model with this aim in mind, scaling training and inference with powerful NVIDIA GPUs allows Haiper to open the door to further use cases and contexts.

Haiper was founded in late 2021 by Dr Yishu Miao (CEO) and Dr Ziyu Wang (CTO), both with PhDs in Machine Learning from Oxford University and former Researchers at DeepMind. Dr Wang was a key contributor to DeepMind programs AlphaStar and AlphaGo and also served as a Staff Research Scientist at Google Brain, while Dr Miao previously founded and led the London ML team building large language models (LLMs) for Global Trust & Safety at TikTok. The company’s co-founders have worked alongside AI pioneers, including Geoffrey Hinton, Nando de Freitas (principal scientist of DeepMind), and Phil Blunsom - currently Chief Scientist at Cohere. Haiper has also partnered with several top academic labs from the likes of the University of Oxford, University of Cambridge University of British Columbia, Simon Fraser University, and Fashion Innovation Agency at London College of Fashion.

Dr Yishu Miao, co-founder and CEO of Haiper, comments: “This collaboration marks a groundbreaking exchange of knowledge and expertise. Haiper is looking to help revolutionise the generative AI landscape through advanced research and development. Our work  with NVIDIA marks a watershed moment as we boldly forge ahead into the cutting-edge future of AI technology.”