Evidence receipt / evaluation
Published · transcript-backedComfyanonymous: evaluation
4 Jan 2025 Latent Space AI Engineering for Art — with comfyanonymous, of ComfyUI
“The reason for that approach was because basically they had two models and then they wanted to publish both of them.”
Source trail
Everything needed to verify it.
- Speaker
- Comfyanonymous
- Attribution
- Verified speaker
- Claim type
- evaluation
- Recorded
- 4 Jan 2025
- Publisher
- Latent Space
Transcript context
…But they didn't, they didn't pursue it for like SD3. What do you mean? Like the SDXL approach. Yeah. The reason for that approach was because basically they had two models and then they wanted to publish both of them. So they, they trained one on. Lower time steps, which was the refiner model. And then they, the first one was trained normally. And then they went during their test, they realized, oh, like if we string these models together are like quality increases. So let's publish that. It worked. Yeah. But like right now, I don't think many people actually use the refiner anymore, even though it is actually a full diffusion model. Like you can use it on its own. And it's going to generate images. I don't think anyone, people have mostly forgotten about it. But, uh. Can we talk about models a little bit? So stable diffusion, obviously is the most known. I know flux has gotten a lot of traction. Are there any underrated models that people should use more or what's the state of the union?…
Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.