4 ms·
> Veo's cutting-edge latent diffusion transformers reduce the appearance of these inconsistencies, keeping characters, objects and styles in place, as they woul
by miohtama 2y ago
> Veo's cutting-edge latent diffusion transformers reduce the appearance of these inconsistencies, keeping characters, objects and styles in place, as they would in real life.
How is this achieved? Is there temporal memory between frames?
- hackerlight 2y agoProbably similar to Sora, a patchified vision transformer, you sample a 3d patch (third dimension is time) instead of a 2d patch