![Flux LoRA/GGUF Minimal Workflow [Simple Beginner]版本v2 LoRA (ID: 733970)](https://image.1111down.com/xG1nkqKTMzGDvpLrqFT7WA/418499b4-fbed-4299-8886-fb66fc3939ad/width=450/24775057.jpeg)
I highly recommend using v1 GGUF at this time:
GGUF is a great way to save time from what I've experienced, only requiring about 6 steps rather than 20. This workflow doesn't include a LoRA node, but one could easily be added, this is mostly to try to help beginners know there's a more efficient way to generate. The results are comparable to the base Flux D model, although not always as good.
Using a cfg of 1.0 produces the fastest results. It takes about 9 seconds per generation with my 4070 Ti Super at about 1.5s/it for a 1024x1024 image.
The default uses 0v GGUF Q4 (UNET) https://civitai.com/models/630820?modelVersionId=736086
v2:
This a simplistic version of a flux workflow which uses a LoRA in ComfyUI using separate nodes for each stage of the workflow.
Updated to use the even simpler version of Save Image (Extended) which also allows for dragging and dropping .jpg files directly into ComfyUI. The reason being is that .jpg are much smaller on average than .png and most AI images don't include empty space.
"This custom node is largely identical to the usual Save Image but allows saving images also in JPEG and WEBP formats, the latter with both lossless and lossy compression. Metadata is embedded in the images as usual, and the resulting images can be used to load a workflow." by palant
The default uses "FLUX ?FaeTastic Details" LoRA which can be found here: https://civitai.com/models/643886/flux-faetastic-details
描述:
Updated to use the even simpler version of Save Image (Extended) which also allows for dragging and dropping .jpg files directly into ComfyUI. The reason being is that .jpg are much smaller on average than .png and most AI images don't include empty space.
"This custom node is largely identical to the usual Save Image but allows saving images also in JPEG and WEBP formats, the latter with both lossless and lossy compression. Metadata is embedded in the images as usual, and the resulting images can be used to load a workflow."
by palant
训练词语:
名称: fluxLoraGGUFMinimal_v2Lora.zip
大小 (KB): 2
类型: Archive
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success