tech
Google’s new anything-to-anything AI model is wild
Posts from this topic will be added to your daily email digest and your homepage feed.
TL;DR
- Omni is Google's new family of generative models capable of turning various inputs into other media, initially focusing on video generation.
- Omni Flash, the first released model, is available on Google's Flow platform and improves upon the previous Veo model.
- Omni allows users to use uploaded videos and text prompts as starting points for AI-generated content, promising better consistency and real-world knowledge.
- Testing revealed mixed results, with some clips showing significant improvement and consistency, while others contained AI 'jump scares' and object inconsistencies.
- Editing capabilities have improved with Omni, but results can still be flawed, sometimes requiring users to generate new videos from scratch.
- Deepfake experiments of the author yielded convincing results, with some videos being difficult to distinguish from reality upon first viewing.
- Generating videos and edits with Omni requires credits, which can accumulate costs, especially for users with specific or complex visions.
- Despite advancements, the author concludes that while Omni is a significant step forward, the technology is still within the 'uncanny valley' and not yet at the level of the 'singularity'.