The evolution of AI video platforms has been rapid over the last few months. While AI video has struggled with consistency and movement in the past, a number of new companies and improvements from existing companies appears to have solved many of these challenges. While it’s still not perfect, AI video is now capable of creating very convincing amalgamations.
This was all before Google shocked everyone with the reveal of Veo 3, a text-to-video platform that is capable of multi-modal productions that include speech with lipsyncing, sound effects, and music to provide a full scene experience. Veo 3 is the first platform to really be able to handle text-to-video consistently.
Several AI video generator platforms are capable of text-to-video. Still, they are focused on camera and people movement as well as mise-en-scène, and less on speech and audio. While these platforms are impressive, they still struggle with consistency when trying to generate a video from text only. However, when giving these platforms a starting image to work from, as well as a prompt of what you are looking to achieve, the results are impressive.
Each platform offers its strengths and weaknesses, but the best way to see this is by putting several of the leading platforms through a few tests. These tests need to showcase how they handle camera movements, especially when moving beyond the content within the created image. Also, tracking how it can handle a consistently moving character in unnatural light. Finally, evaluating fast tracking and movement.
Three images have been created using MidJourney, and then using these images as a base, prompts have been written up so all platforms have the same instructions to ensure a fair comparison. Veo 3 has also been given the same video prompt, without the base image, to see what it can come up with. There are also three examples at the end of Veo 3 containing dialogue as well to show more of its capabilities.
Camera Movement
Prompt: “A slow, cinematic drone shot flying over an eerie, fog-covered abandoned amusement park. The camera glides above rusted rollercoaster tracks, broken-down rides, and an overgrown carousel. The Ferris wheel creaks as it turns slightly on its own. Birds scatter as the drone approaches. The lighting is cold and desaturated, with light fog creating a mysterious, haunting atmosphere.”
People and Lighting
Prompt: “A cinematic night scene in a futuristic city, neon lights reflecting. The woman turns her head and looks directly into the camera. As she turns, the camera begins to zoom in smoothly on her face, revealing her face and raindrops trailing down her cheeks. Neon signage flickers around her, steam rises from vents nearby, and the rainfall continues steadily, adding a soft ambient rhythm. The mood is quiet, intimate, and slightly tense, like a key moment in a sci-fi drama.”
Action/Fast Movement
Prompt: “1970s muscle car in the heart of gritty New York City. The car suddenly accelerates, racing past the camera. As it speeds off, the camera quickly swings around and follows close behind in a chase-style tracking shot. The car weaves through traffic, barrels down narrow alleys, and skids around street corners, tires screeching. Storefront lights and signs blur as it speeds through puddles, splashing water under the glow of streetlamps. The streets are alive with honking cabs, pedestrians jumping out of the way, and the hum of an old-school funk or soul soundtrack. The tone is high-energy, cinematic, and reminiscent of classic 70s car chase scenes.”