How long is generating with txt2img? Correct way to generate?

#162
by PufferDuck - opened

I've set up a local instance and experimenting with some prompts. Generation is taking a while.
I am just using txt2img without modifying any parameters. It takes some time to load a model (sd-v1-4-full-ema.ckpt) that I've placed in models\ldm\stable-diffusion-v1\model.ckpt and around 40 seconds to generate.
GPU is RTX 3090.

Is it expected time to run it?

And is it correct way to approach generation? I also downloaded all the models via script, but can't figure out how to try them, mind helping me find a documentation?

Let's try autocast.
you can get a fast image generation.

Sign up or log in to comment