We’re wildly excited to present our most recent AI Short Film at The Awarehouse; H3O2!
What a strange name for a film. But we live in a strange world.
Water has long been a topic around our fire at the House. Not only the otherworldly nature of the substance and the small fact that all life depend on it, but the highly mysterious action going on at the molecular level, not the least; water memory. We invite you on this short journey upstream where we ponder some things that might ponder us right back.
H3O2 is a visual poem tracing the journey of a single water droplet as it forms in the clouds and sets off across the world. Each encounter leaves a trace, impressions held in the fluid memory of its form. It finally returns to the sea where it discovers its part in a larger whole and its contribution to the tapestry of life.
We also want to send a virtual high-five to Curious Refuge and PromiseAI for rolling out the world’s first AI Animation Competition! If you haven’t already, check out the submissions on their page (curiousrefuge.com), it’s amazing to see the films that people create using these tools!
This project was so much fun to create, and it wasn’t that technically hard! Let’s break it down;
We used krea with a bunch of character images to train a lora on the droplet character. We did the same with the style by uploading a bunch of images in the style we wanted. Now, here’s a tip in regards to that; Just as you want different poses and expressions from your character, you also want variety in your style references. The model is training on more than style as we usually talk about it, i.e. ghibli, sketch, pixar or whatever, it is influenced by the composition, the depth of field, the lighting, amount of characters in the etc. So it’s worth doing some extra (even manual) work on those shots to match the vision you have.
Once we had our character and style set, we worked between Krea and Canva Whiteboard to set up our story in images. Simultaneously we run images in Hailuo ai with an unlimited plan to get a rough draft out on an editing timeline (Premiere Pro), so that we can begin to shape the soundscape (using Artlist, Udio and Elevenlabs) while finalising our story with images on Canva. To get consistency in the images we use a lot of photoshop, and sometimes run a clip in Hailuo AI Director mode to capture a screenshot from the scene at a different angle – and into Magnific.
When that is done we run all the images through Magnific and finally Kling with more intention in the prompt for acting and camera movement. Then we baked it all together in Premiere. For the VFX we used After Effects to layer in the memories in the water, we used Meshy AI to create the 3D character, Mixamo to get the animation of it, and Unreal Engine to put together the shot when it falls down the drain. And finally, upscaling the video in Topaz.
Thanks for dropping by!
