| ▲ | n2d4 17 hours ago | |
Source? It's much more likely that the LLM generates the latent vector which serves as an input to the diffusion model. | ||
| ▲ | jumploops 16 hours ago | parent | next [-] | |
From the GPT-4o System Card Addendum[0]: > Unlike DALL·E, which operates as a diffusion model, 4o image generation is an autoregressive model natively embedded within ChatGPT. [0]https://cdn.openai.com/11998be9-5319-4302-bfbf-1167e093f1fb/... | ||
| ▲ | og_kalu 16 hours ago | parent | prev [-] | |
Open AI said it's auto-regressive, the presentation on the app is autoregressive, it's priced auto-regressively. Why would that be more likely ? It seems like some implementation of bytedance's VAR. | ||