The short film about the rainforest created entirely with AI – are you fooled?

Creating images with AI and so-called "promptography" is no problem for most people – for example, through Adobe Firefly, Midjourney or Stable Diffusion.

However, creating high-quality and detailed video material with the help of AI is much more laborious. The models that generate the images consume a lot of processing power, and it also takes time to render sequences to see the results and then redo with new changes and improvements.

"This is a project I started to test out the SVD weights in Stable Diffusion (Comfy). Visually, everything is locally rendered on my machine (RTX 3090Ti) using a variety of different tools and software for image generation, image animation, frame interpolation, and post upscaling (I found it better than running downscale to upscale hiresfix - too much detail was lost)."

The model produced sequences at a resolution of 1536x640 at 8 frames per second, which was then interpolated up to 3840 x 1620 at 48 frames per second. The voiceover was created with Chat GPT.

Also check out the clip below from ActionMovieDad on X, who writes about his clip created in Midjourney:

"This literally started with me seeing if the new Midjourney could do hands, then hands in water, and it spun out of control. I promise this is as hard as I’ll ever lean into that classic Midjourney male gaze - but also I kinda like how it turned out."

For instance, check out the demo Matrix Awakens and compare it with this rainforest sequence.