1. Haiper is a new artificial intelligence startup specializing in generative video models, aiming for Sora-like realism in the future.
2. The company is focused on quality, offering short two-second clips with intuitive motion capabilities, but struggles with animating larger objects.
3. Haiper’s AI model is better at anticipating motion with shorter prompts, making it an impressive addition to the AI video market with potential for growth and improvement.
Haiper is a new artificial intelligence startup focusing on generative video models, aiming for realism similar to OpenAI’s Sora. Founded by former Google DeepMind and TikTok engineers, Haiper is free and aims to democratize creativity. The underlying model excels at determining motion in short clips without extensive prompting.
In testing Haiper with various video clip ideas, it performed well with small motions and animated still images effectively. However, it struggled with larger objects like animals and people, often merging or blurring details. The AI excelled at creating realistic visuals but had difficulty with consistent motion across the clips.
Despite its limitations, Haiper shows promise in the AI video market and is expected to improve with time. While it currently falls short of models like Sora in longevity and motion consistency, the company’s recent funding round suggests growth and advancement in the near future. With a focus on simplicity and improved training data, Haiper aims to become a competitive player in the AI storytelling space.