AI-powered Video Generation Startup Pika Founder not Intimidated by OpenAI's...

BEIJING, February 20 (TMTPOST)—OpenAI's recent release of Sora has taken the world by storm. Functioning as an AI video model, Sora has the capability to create realistic and imaginative scenes based on textual instructions. A user describes a desired scene in words and Sora will return a high-definition video clip.

Sora takes language comprehension to a new level, accurately understanding cue words and producing emotionally vibrant videos. Built upon comprehensive research into models like DALL-E and GPT, Sora introduces a new model paradigm. It not only interprets user requests from prompts but also understands their presence in the physical world.

In particular, Sora is a diffusion transformer, with transformers demonstrating outstanding scalability in various domains such as language modeling, computer vision, and image generation.

As a diffusion model, Sora goes beyond generating videos based on textual instructions; it can extract existing static images and create videos from them, animate images accurately, and pay attention to details. Sora can also generate video clips inspired by still images, and extend existing videos or fill in missing frames.

According to OpenAI's technical report, the research outcomes of Sora indicate that expanding video generation models is a highly promising approach to building a universal simulator for the physical world. It enables artificial intelligence to understand and simulate the physical world's movements, marking a new milestone in the journey towards achieving Artificial General Intelligence (AGI), not limited to video generation.

Before Sora was introduced, Runway and Pika were considered frontrunners in the video generation field. With Sora's release, many believe it has effortlessly outperformed and is poised to outshine these two emerging unicorns, causing concerns for entrepreneurs in the related sector.

However, Pika seems more exhilarated than fearful about the situation. In a response to TMTPost’s request for comments, Guo Wenjing, the founder of Pika, said: "We find this news very exciting, and we are preparing to benchmark against Sora."

Furthermore, Guo revealed that Pika is currently hiring, with specific plans yet to be disclosed.

Established in April 2023, Pika Labs released its first product, Pika 1.0, in November 2023. Pika 1.0 can generate and edit 3D animations, cartoons, and movies. Ordinary users can also modify the content, positioning it as a user-friendly "video generation tool."

"I think videos are different from language models. We already have a general idea of the methods for language models, and the algorithms are quite mature. But there is no well-grounded algorithm for videos at the moment. It's not a scalable problem; it's not that people don't have enough GPUs now, it's because there isn't a good approach algorithmically," Guo explained.

Sora's launch now provides the industry with a solid algorithmic approach, potentially offering more mature algorithmic pathways for leading AI companies like Pika.

In fact, Guo has made full preparations to compete against OpenAI. As a female growing up in East Asia, Guo graduated from Harvard University with a bachelor’s degree and dropped out of Stanford's Ph.D. program to embark on entrepreneurship. Pika 1.0's demo video garnered instant attention for its ability to generate and edit various video styles, and its user-friendly interface makes it accessible to everyone.

Pika Labs, founded by a team of four people, raised over $55 million in funding, with participation from almost all prominent early-stage investment firms in the AI domain.

Just four months after the release of Pika 1.0, Sora's sudden debut adds more uncertainties and possibilities to the AI video generation landscape.

In an exclusive interview with TMTPost, Guo talked about her interest in the creative industry and how AI leads her to explore ways of using it in video production.

She highlights the challenges in traditional video production, such as complex editing processes and the time-consuming nature of certain tasks. Guo participated in Runway's AI video competition, where she realized the limitations of existing tools and saw an opportunity to address these challenges through her own venture, Pika.

Differentiating Pika from competitors like Runway, Guo emphasized the focus on making AI-powered video creation tools accessible to ordinary users, not just professionals. She believed in the significance of the technology and the need for innovative approaches in video editing.

Addressing the issue of data, Guo acknowledged the competitive landscape but emphasizes that superior technology remains a crucial factor. She pointed out that Pika's team comprises top talents from prestigious institutions, and their strategic positioning targets a broader user base than some competitors.

Regarding the challenges in AI video production, Guo mentioned algorithmic limitations, particularly in generating long and high-quality videos. She underscored the need for breakthroughs in algorithms to overcome these limitations.

When it comes to future plans, Guo envisioned Pika as a tool for ordinary users to create and edit videos, rather than a platform like TikTok. She emphasized their commitment to breaking through technical bottlenecks and maintaining an open attitude towards user applications.

In response to the evolving AI landscape, Guo expressed confidence in the ongoing AI revolution, noting that AI will undoubtedly transform various industries and applications. She believe the potential of AI to disrupt traditional methods, including video and music production and shopping.

Guo also reiterated the dynamic nature of the AI industry, the importance of a skilled team, and the continuous pursuit of technical advancements to meet the challenges and opportunities in the evolving landscape.