|
|
|
|
|
|
by vunderba
624 days ago
|
|
|
The answer is it really depends on your hardware, but the nice thing is that you can split out the text encoder when using ComfyUI. On a 24gb VRAM card I can run the Q8_0 GGUF version of flux-dev with the T5 FP16 text encoder. The Q8_0 gguf version in particular has very little visual difference from the original fp16 models. A 1024x1024 image takes about 15 seconds to generate. |
|
|