you like my content, support me at:<a target="_blank" rel="ugc" href="https://ko-fi.com/duchaiten">https://ko-fi.com/duchaiten</a>Patreon in here: <a target="_blank" rel="ugc" href="https://linktr.ee/Duc_Haiten￼￼I">https://linktr.ee/Duc_Haiten I</a> created my discord server. I'm still quite new to discord server management, free to give opinions.<a target="_blank" rel="ugc" href="https://discord.gg/FPhQvFsqe2">https://discord.gg/FPhQvFsqe2</a><hr />The same dataset trained on Dev and Schnell seems to perform better on Dev, as it shows fewer anatomical errors, more accurate colors without being overly saturated, and sharper, more precise details. It is also more flexible and diverse with the same prompt and random seed. However, it trains slower than Schnell, has less artistic quality, and doesn't learn as closely to the training data as Schnell does. Sometimes Schnell appears to adhere to the prompt more strictly, and I think this might be because Schnell works better with a lower CFG than Dev, which causes this phenomenon. Currently, there are ways to increase the true CFG for Dev, but it’s said to increase image generation time, and so far, this can only be done on ComfyUI.

Files with SHA256: 472946ca51a7199cc2d492ead361ec25004f8a6b4282d31bd200c7f1920d6572