NEW BEST AI MODEL! How to use Flux.

[Puckerization]

You can now add the models to the Checkpoints folder in the latest version of ComfyUI. Change the .sft extension to .safetensors

[IIStaffyII]

Was waiting for this video as soon as i read about flux!

[mlegarth]

Latest update actually breaks this. Flux was running great on my 4090 until this morning. Now the clip models don't load properly and generating time is insane. From 1.5 it/s to 100 s/it.

[MrLogansrun]

Thanks for the vid buddy will try this later after work 👍

[jakobmayr]

Unfortunately i get an "list index out of range" error...any tips ?

[NotThatOlivia]

test other scheduler + sampler combos - some of them are really nice

[inglesconpelis]

I don't know much about pcs but I am on a 12th generation core i5 16 gb ram and envidia 3050 with 4gb of vram I am thinking about getting an external GPU like 4070 nvidia but I am wondering if that might be useless for running a model like flux because of my cpu does anyone know if cpu like mine will be limites for this?

[TheGHOST]

I've been eagerly anticipating your video on Flux... greetings from Italy! I've been experimenting with Flux for a few days now, and it's been a game changer! 😂😂😂😂

[TomiTom]

Thanks for the tutorial Sebastian as always great work.

BTW I am using FLUX in a totally different UI, called SwarmUI. This tool has supported the FLUX model immediately, and I can generate an image for Schnell model only in about 40 seconds, and for dev model in 3 minutes. I have RTX 3070 Ti with 8gb vRam. While in ComfyUI it takes 3 minutes for the schnell model.

SwarmUI is the UI that will replace A1111, it is so optimized and well organized, I just hope you try it, Sebastian, and if you find it any good to cover it in your next video. I found very few channels who mentioned it.

[JohnToridas]

used it on nvidia 3060 12GB and ryzen 3100 16RAM, at first loads like 20min, and then 5 min for generation, quality outstanding

[TheOriginalOrkdoop]

Will flux be available on a1111? Im afraid to try comfyui bc im used to a1111.

[MarkJoel]

What... no dad joke? j/k -- nice job as always.

[soulspin-project]

Sa upp ideogram igår. Får typ samma resultat med Flux. Riktig bra model.

[zuriel]

Whoa, I had no idea that it could be ran with an 8gb GPU. I heard stories of people saying it took them over 2 mins per gen with a 3090 so I just completely dismissed ever trying since I only have a 3070. Maybe I need to give it a try myself and see what happens

[darkman]

How many workflows are there and which one/s are best for flux? How do you know which one to choose?

[axel]

Is it possible to use flux on fooocus?

[laser__unicorn]

So which one is best if my GPU is 8GB but my RAM is 32GB? Would using the fp16 cause any issues/errors or is it better to stick with fp8?

[MrLogansrun]

I have a rx7800xt struggled to get comfyui working on it. So I put a rtx3060 with 12gb in with the rx and used the cuda version . Wonder if there is a way i could use the 16gb off my rx and the 12 on my rtx together? It seems that it will only use the rtx 3060 . Maybe one day they will implement amd better .

[GarethOakey]

Where should you add the controlnet node?

[Antilopax]

lol this Video is a couple of hours old and already some files have different names.. Crazy how fast this space develops