New open-source AI video generator is INSANE
Science & Technology
Introduction
Ladies and gentlemen, the moment we've all been waiting for is finally here! We now have an open-source AI video generator that boasts Sora-level quality. The tool is called Pyramidal Flow, or Pyramid Flow for short, and it’s developed by the same company that created Cling. The fact that they are open-sourcing this is simply incredible!
Demonstrations and Features
Pyramid Flow can generate 10-second clips at 24 frames per second and a resolution of 1280 x 768. Here are some demonstrations of what it can do:
Profile Shot with Fireworks: A side profile shot of a woman with fireworks exploding in the distance showcases its impressive capability. In comparison, a similar generation from Runway Gen 3 only highlights the advancements made in Pyramid Flow.
Kebab Close-Up: An extreme close-up of chicken and green pepper kebabs grilling on a barbecue is generated with realistic flames, shallow focus, and vivid colors.
Busy Snowy Tokyo: A bustling Tokyo city scene delightfully captures the snowy atmosphere. Although some edges of pedestrians may warp over time, the overall temporal consistency puts it ahead of previous open-source video generators.
Highway Scenic View: Another spectacular generation shows a car driving on a highway at dusk with a scenic sunset reflected in the rearview mirror.
The fidelity continues with fantastically crafted scenes, including:
Space Adventure Trailer: A cinematic video depicts The Adventures of a 30-year-old spaceman in a vibrant, colorful setting.
Leisurely Boat Ride: A black and white clip of a boat sailing along the San River, offering impressive consistency.
Cat waking Up: The detail on the cat’s fur and pillow realism in this clip is mind-blowing.
Landscapes: Shots like waves crashing against rugged cliffs are rendered beautifully and show almost no flaws.
Moreover, this tool can cater to intricate scenes even with multiple characters and objects, greatly improving upon prior generations of video generators.
Performance and Accessibility
What sets Pyramid Flow apart from its predecessors is its relatively lower computational requirements. The open-source model is meant to run without requiring immense GPU capabilities compared to Sora, which claims high computing demands. For those wondering about specs, Pyramid Flow reportedly requires around 26 GB of VRAM for 380p resolution and up to 40 GB for 768p resolution, meaning even powerful GPUs like the RTX 4090 might struggle.
Image to Video Capabilities
Further enhancing its versatility, Pyramid Flow can also create videos from images. Users can provide a starting image and a prompt, leading to appearances like cars driving down the road or animated sequences inspired by classic paintings.
Benchmark Performance
Across various benchmark metrics, Pyramid Flow has achieved a total score of 81.7, outperforming previous generations of closed-source video generators like Runway Gen 2. Its quality is also notably superior to competing tools.
Conclusion
The open-source community is buzzing with excitement about Pyramid Flow, and its potential applications in creative projects are expansive. For those interested in experimenting with it, the developers have already released the code and model on GitHub.
Keywords
Pyramidal Flow, AI video generator, open-source, Sora level quality, video demonstrations, computational requirements, image to video capabilities, benchmark performance, realistic video generation, Cling.
FAQ
Q: What is Pyramid Flow?
A: Pyramid Flow is a new open-source AI video generator developed by the same team behind Cling.
Q: What kind of clips can Pyramid Flow generate?
A: Pyramid Flow can generate 10-second video clips at 24 fps with a resolution of 1280 x 768.
Q: How does Pyramid Flow compare to other AI video generators?
A: Pyramid Flow outperforms previous generations, both open-source and closed-source, in quality and temporal consistency.
Q: What are the hardware requirements for running Pyramid Flow?
A: To run the 768p version of Pyramid Flow, you will need approximately 40 GB of VRAM, while the 380p version requires around 26 GB.
Q: Does Pyramid Flow have image to video capabilities?
A: Yes, Pyramid Flow can create animated videos from static images, making it versatile for various applications.