generated this image on my m2 mac with 32gb of ram in about 38 seconds:
using:
https://github.com/filipstrand/mflux + schnell model
impressed... could replace my use of midjourney.
nostr:note1pnezaa9ehx0kwt5pcxn33vrx27pa2uxsakusudvrqsj42tg7nd2szwxg7e
Adding this to my library of nostr school lessons. 🤣
I've been using stable diffusion locally for a few months on a recent AMD laptop GPU. Works great!
no need to ever pay subscriptions for generative ai
hugging face is the shit 🔥
Looks real to me.
Would you mind sharing your prompt?
was just the example prompt in the repo:
mflux-generate --model schnell --prompt "Luxury food photograph" --steps 2 --seed 2 -q 8
Ah gotcha. I was just asking so that I could see if I could obtain the same realism on my Ubuntu desktop box with a 7900 XT GPU. Don't have a flux-based model currently setup though.
Via comfyUi or with the mentioned way?
there is https://github.com/raysers/Mflux-ComfyUI but I haven't tried comfyui yet
damn... that fooled me
in order to run mflux, I had to update python. In order to update python, I had to install brew. Uh!
did you encounter this issue?
-access to model black-forest-labs/FLUX.1-dev is restricted.
you need to accept the terms and conditions on huggingface while logged in
Thanks, I am able to run the following script now but it has been running for 5 mins now. Not settled yet. M3 chip 18 GB.
--model dev --prompt "Luxury food photograph" --steps 25 --seed 2 -q 8
I've never had the need for image generation, but I feel tempted to try this anyway on my M4 :-)
Curious how fast it is, do you have an m4 max ?
A mere Pro
With 48GB RAM. The food benchmark takes 38 seconds, but towards the end it uses a bunch of swap. I have about 6 GB of other stuff eating my RAM and didn't try pruning that yet.
With quantize 8 it fits in memory and takes 33 seconds.
"Death Star construction" with the default 4 steps took a minute. And no audible fans. 
decent... im impressed that these few-step models work so well. the dev model takes like 20-25 diffusion iterations and is much slower... I haven't noticed a huge quality difference.
