1. Runway Gen-3

For those who may not be aware, Runway Gen-3 Alpha is now accessible to the public.

The New York City-based firm Runway has made a notable return after a year since the launch of Gen-2. Gen-3 Alpha introduces a new line of models developed on advanced infrastructure tailored for extensive multimodal training. This version demonstrates remarkable enhancements in fidelity, consistency, and motion when compared to its predecessor.

Here are a few illustrative examples:

> Prompt: Subtle reflections of a woman on the window of a train moving at hyper-speed in a Japanese city.

This example highlights Gen-3 Alpha's proficiency in rendering intricate reflections and swiftly moving subjects with outstanding realism.

> Prompt: An astronaut running through an alley in Rio de Janeiro.

The model's capability to create intricate environments and realistic human motions is clearly visible here.

Cost:

The subscription fee is set at $15 per month, or $12 per month if paid annually.

For more information about Runway Gen-3, visit their site.

2. Kling

Kling, the latest AI video generator from Kuaishou (which translates to "quick hand"), a Beijing-based competitor to TikTok, has arrived.

Kling can produce videos up to 120 seconds long at 30 frames per second with 1080P resolution and a flexible aspect ratio. Its developers claim that this AI model has a superior understanding of physics, enabling it to simulate complex movements accurately.

Check out this example video:

> Prompt: A Chinese man sitting at a table, eating noodles with chopsticks.

> Prompt: A man riding a horse in the Gobi Desert, with a beautiful sunset behind him, a movie-quality scene.

This showcases the impressive temporal coherence of Kling’s outputs.

Access Information:

Currently, Kling's AI model or app is not widely available. It is reportedly accessible via the Kwaiying app for select beta testers.

For the latest updates on availability, check their official website, although it is primarily in Chinese.

3. Vidu

Vidu, a text-to-video AI model crafted by ShengShu Technology in collaboration with Tsinghua University, was introduced on April 27, 2024. This model is designed to generate high-definition, 16-second videos in 1080p resolution with just one click.

According to Zhu Jun, the chief scientist at Shengshu:

> "It is imaginative, can simulate the physical world, and produces 16-second videos with consistent characters, scenes, and timelines."

Here’s a demonstration:

Access Details:

Vidu is not yet available for the general public. However, they have launched a waitlist for early access:

1. Visit www.shengshu-ai.com
2. Click the blue button located at the top right of the page
3. Complete the form to request access

Additionally, plans are underway to integrate video generation capabilities into an AI tool named PixWeaver.

4. Google Veo

Veo represents Google’s most advanced model for video generation, capable of producing high-quality videos in 1080p resolution that exceed one minute. It accommodates various cinematic styles and accurately interprets prompts to capture nuanced details.

Veo builds upon years of research with generative video models such as Generative Query Network (GQN), DVD-GAN, Imagen-Video, Phenaki, WALT, VideoPoet, and Lumiere, along with the Transformer architecture and Gemini.

To enhance Veo’s understanding of prompts, more detailed captions have been included in its training data. Improvements have also been made using high-quality, compressed video representations, allowing for faster and higher-quality video generation.

Key Features of Veo:

• Generates coherent scenes by combining text prompts with visual references.
• Edits videos based on specific commands and masked areas.
• Utilizes reference images to guide video creation.
• Extends video clips to 60 seconds or longer from single or multiple prompts.
• Maintains visual consistency across frames with latent diffusion transformers.

Take a look at this example:

> Prompt: A fast-tracking shot through a bustling dystopian sprawl with bright neon signs, flying cars, mist, and volumetric lighting.

Interestingly, none of the example videos presented by Google feature clear human faces, focusing instead on animals, environmental scenes, or floral imagery. It remains uncertain when Google will make this video model publicly accessible, although it is anticipated to be integrated into their AI chatbot, Gemini.

Conclusion

It's exciting to observe the advancements in AI video technology catching up with text and images. Although many of these tools are not yet publicly accessible, their preview outcomes are impressive. Explore these tools to determine which best suits your requirements and budget.

Stay vigilant for these innovations, and try them when possible to see which ones align with your objectives.

Regarding the impact on the job market, it seems unlikely that pure text-to-video technology will soon dominate the film industry. Rather, video-to-video technology might emerge as the primary tool for filmmakers, at least until AI models can process entire narratives and produce complete films in one go.

What are your thoughts on these AI video generators?

This article is published on Generative AI. Connect with us on LinkedIn and follow Zeniteq to stay updated on the latest AI developments.

Subscribe to our newsletter for the latest news and updates on generative AI. Let's shape the future of AI together!