I recently had the privilege to be a judge in Project Odyssey’s AI film competition, and I am blown away by how far generative AI has come in such a short time. One thing that stood out to me was that nearly every entry used a blend of conventional filmmaking tools, such as Final Cut and Premiere, augmented by generative AI tools of all modalities, closed and open source. I know when Stable Diffusion 1.4 dropped just over 2 years ago, the thing that had me most excited was how it was going to empower indie creators. The technology has come a long way since then, as has our collective skill in using it, and with the work on display at Project Odyssey, we can increasingly see the positive impact it is having on indie creators. We are fast approaching a day when anyone with a story to tell will be able to tell it and see their vision fully realized, even if all they have is an old laptop and their phone camera.
ComfyUI, a powerful node-based graphical user interface for building generative AI workflows, seemed to be nearly ubiquitous in the creation of everything I viewed. This implies creators were frequently running AI models like Flux and LTX Video locally on their own computers, and not solely relying on cloud services like Kling, Runway, or Midjourney. Within the “Rendering and VFX” category especially, ComfyUI is a great tool for applying consistent rendering effects to huge numbers of frames. Some films submitted were almost 15 minutes long, which is 21,600 frames at 24 frames per second. It also has integrations with popular digital art tools such as Krita, allowing artists to blend AI into familiar tools and workflows.
ElevenLabs‘ incredibly lifelike AI synthetic voices also featured prominently in many creations, but there are impressive innovations happening in open text-to-speech models as well (Kokoro-82M anyone!?), nearly all of which can be easily run on home computers and consumer-grade GPUs. The same can be said in music generation, currently dominated by services like Udio and Suno. I would expect to see a lot more development in open music generation models in the next year, and Project Odyssey Season 3 may use many more open tools for speech and audio generation.
There’s been an explosion in high quality video generation models recently, including LTX Video (published December 2024) from Lightricks and Cosmos (published January 2025) from Nvidia, but even older techniques like AnimateDiff (published in July 2023) found their way into many of the best entries. It’s a rapidly evolving modality, with new models released almost every week. The outputs of these models are often clips no longer than 10 seconds, so the use of traditional video editing software and skills is still very much required to produce long-form content. These models also often generate videos at fairly low resolutions, so it makes sense that so many entries relied on video upscaling services like Topaz. There are open AI upscaling models, but none of them are quite at the level of Topaz yet.
3D model generation is a modality that has seen some critical advancements in the last year, and several entries utilized AI text-to-3d models alongside traditional 3D tools like Blender in a hybrid workflow, ultimately producing the final video with traditional 3D animation and rendering techniques. Though not the subject of this competition, I would expect to see an impact in indie video game development from this as well.
For me personally, and for everyone at Salad, it’s the explosion in excellent open-source tools and open-weight models that run on consumer GPUs that has us most excited. Salad loves open source, and we build in public! SaladCloud is the most affordable, most scalable, and most flexible solution for running image and video generation workflows, with customers running thousands of simultaneous GPUs, scaling up and down on demand. Generative AI moves at an absolutely breakneck speed, and Salad has the resources and expertise to help you keep up and scale.
A special shoutout to some of my favorite entries, in no particular order:
- The Myth – Animated Series
- ANPU
- The Kitchen Tapes: Mr. Garlic’s Last Session
- KITSUNE
- The Ideal Future Sanctuary
Additional Resources:
