Видео

yeah it took me a while to figure out what i needed. ended up returning and reordering a few things. worked out pretty well though!

the mobo has 4 m2 slots on it. idk about the card.

@@flowstateeng Okay, thanks. So far I’m really trying to save as much money as possible on this build as a budget workstation. I’m looking at the 2565X and probably just 4x16Gb or 4x32GB DDR5 RDIMM. I will need a waterblock as well. But other than that, I think I’m covered. What went wrong with your build? How was ram support?

i didnt read the mobo cpu support list. with consumer mobos, the socket determines the cpu, so i saw the mobo supported the lga 4677 socket and got a xeon cpu of that socket type. but with server grade mobos/cpus, there is more to consider. just because its the same socket, doesnt mean the cpu will be supported. so i ended up needing to order a different cpu. and unfortunately, they didnt let me return the old one, so i paid for 2 cpus. but i guess now i have an extra lol

yep, thats exactly what i was thinking. im working on some more controlnet and ipadapter vids now doing exactly that!

yeah im sure there are some that have decent effects. for now, im happy with flux + controlnet/ipadapter!

yep, i have that one. it just doesnt show in the new ui for some reason. i saw one guy who had it in the top menu bar in the new ui but i havent pursued figuring out how to do that yet

well i just updated comfy and now those meters are showing in my manager bar in the new ui. guess thats all it took!

oh yeah it definitely adds complexity! it was a cool experience though. when i add more gpus and radiators, im gonna switch it into an open case so things are more accessible.

yep, im actually working on that now. probably in the next 2-3 days. got a couple other things ahead of it

one way or another, im pretty sure i can get it worked out

Yeah but I don't mind. I really like the FluxSamplerParams sampler anyway. If you haven't heard of that one, check the testing video. That sampler is awesome. ruclips.net/video/EZ9btjSkNR4/видео.htmlsi=OtMiZmb5wVAy5va4

@@flowstateeng yeah i'm using it with a custom version of your own workflow =D it's really good indeed

I tried using mistoline and xlabs depth controlnet with CR multi-controlnet stack node with CR apply multi-controlnet. It worked, but I did not compare this setup with stand alone controlnet setup.

yeah im gonna do some vids on multi control net stacks with this workflow tomorrow. looks really useful

hey yeah i stumbled across that this morning. I'm checking it out now to make sure it works with everything. thanks for the heads up!

the only problem i see is that you lose some of the fine tuning for each preprocessor. like with the standalone canny, you can adjust the low and high thresholds, but with the aux, you dont have that control.

@@flowstateeng from my experience, you can adjust the resolution in aux, which gives similar fine tuning result.

cool ill check it out, thanks!

I love the models lol. I'm toning it down in more recent videos bc some people were complaining that it's a bit distracting, which makes sense. But I'm still throwing a few in there for y'all (and for me lol)

yep, i just made a video about it yesterday. lmk if it helps! ruclips.net/video/COJhS38P-7c/видео.htmlsi=84vC61D-aWuyRh2T

Same here, Searge LLM says null for me as well. Has anyone figured out how to fix this? Thanks.

Hi.. I had this problem too. I didn't find a solution on the internet... after studying solutions and I found one. I created a "llm_gguf" folder inside the comfyui "model" folder and placed the mistral model file there. It worked for me. hope it helps.

That's weird. You created the ""llm_gguf" folder inside ComfyUI/models?

@@flowstateeng yes... and worked!

yep, this is exactly what you're supposed to do. the mistral model has to be in that folder for the searge node to be able to find it. check my searge video for more: ruclips.net/video/46on642Wk3g/видео.html

For sure, no problem! Glad you enjoyed it. I just posted the IP Adapter video and I have several more coming on those and the Control Nets. Those things are both really cool. IMO, much better than LoRAs.

ok my bad i've just watched your videos and workflows that explains it all

Cool, glad it helped! Lmk if you have any other questions!

No problem! Glad you liked it!

check my newer ones. i didnt realize how low it was until someone told me. thanks for the heads up though

I'll eat yo mamma 🎉❤

Hell yeah boooiiiiii

Yeah check this dudes videos. I have my Nvidia settings exactly like his. I should probably make a video on that now that you mention it. You can just ignore the Premiere Pro part of his vid but set all the Nvidia control panel settings like he has it. I'll make a video too that isn't focused on Premiere Pro but for now this might help you ruclips.net/video/K75sc6mKb08/видео.htmlsi=Qp87QA9JBvLdoWMY

Email me (or dm me on X or IG) with a screenshot or screen record

It should only take 2 mins the first time you load it up. Each time you open Comfy and load the workflow, it has to transfer that 23gb model (or 12gb for NF4) from disk to the gpu. So the first generation takes a couple mins. But then if you run it again with a different prompt it should only take 1.5 - 6 seconds per step, depending on your settings. For example, if I do batch size 1 for a 1024x1024 image, it takes about 1.4 sec per step on my rtx 4090. If I do batch size 2 that doubles, or if I do a larger image like 1920x1080, it takes around 2-3 sec per step. What kind of GPU do you have? If its taking longer than you expect every time you run it, you can do a screen record (make sure to show the CMD prompt in it) and send it to me by email or X or IG.

description is much better than title.

Awesome, glad you liked it! That Film VFI node really makes things much better. Yeah I'll revisit the title. I use vidIQ to help come up with good titles but it struggles with some of these. It doesn't know what ComfyUI means really.

absolutely. stable video diffusion only uses about 12gb of vram to generate the video. so the 4070ti super should handle that very efficiently!

Thanks, yeah I've had a few comments about that. Check the latest video and let me know what you think. Should be much better. ruclips.net/video/EgXLKBrnBq0/видео.html

Thanks, yeah I've had a few comments about that. Check the latest video and let me know what you think. Should be much better. ruclips.net/video/EgXLKBrnBq0/видео.html

Thanks! It's taken some practice figuring out what people wanna see but I'm getting things dialed in. the newer Searge and SVD vids have basically no memes. Seems like people don't like them that much, which I get. Let me know what you think if you watch them.

@@flowstateeng relax man, leave the memes. You can cut the amount by 50% but not every ai tutorial on RUclips needs to be the same. I will try your workflow today

Good to know, thank you! I will boost it in the next video!

Thanks! Yeah there are so many options for LLMs in Comfy. At least you have some that work to choose from!

Hi, thanks! I have been considering putting fewer in there. I want to keep it a little lively but it probably could be toned down a little. I'm finding the right balance. I'll post the LLM video today and let me know what you think about that one. Thanks for the input!

Check this one out. I think you'll like this a lotttt better. You were absolutely right about the memes. I like this a lot more. Let me know what you think! ruclips.net/video/46on642Wk3g/видео.html

i honnestly don't even get it why he puts that in the videos.

@thibaudherbert3144 I was thinking that the audience would be really young people with shorter attention spans, so I just wanted to include some action and humor to hold their attention. It seems dumb now, but I didn't realize that the audience would obviously be people who regularly watch various tutorials and enjoy learning, lol. So I've taken them out. Check out the Searge video and the SVD video. There's only a couple of memes in Searge and none in SVD. Hopefully, those are much more enjoyable. Let me know what you think. Feedback is really helpful. Thanks for the input!

@@flowstateeng If you want to buzz and have a lot of views and so targeting a lot of people, you should make cool quick concept using ai in tiktok format videos. But if you are doing tutorials, people stays for the quality/tutorial, not for any other things, so it's pointless in this case 👍

Thanks! The LLM vid is coming either tonight or tomorrow morning. Depends on how late I can force myself to stay up lol

The LLM video is up! Let me know what you think! ruclips.net/video/46on642Wk3g/видео.html

For sure, no problem!

Don't be cry lol

Yes! You absolutely can run Comfy from a portable SSD! I thought your question was very interesting, so I made a video to share with the crowd! Once I post it, I will leave a link here. Thanks for the interesting and useful question! Also, cool username.

Vid is posted here ruclips.net/video/tbPAOl1eaag/видео.html Thanks again for the question, I think people will find it pretty useful!

Haha yeah its a fun learning process! I'm finally dialing in my style, aesthetic, editing process, etc, but it does take some time to get right. Thanks for the sub!

Think I'm getting things dialed in a bit. The LLM video is up and I chilled out with the memes on this one lol. Let me know what you think! ruclips.net/video/46on642Wk3g/видео.html

@@flowstateeng I will have a look into it later. I just start with a buddy together to go into AI picture manipulation and i will try today your tutorial video about setting up a compfyui.

oh nah thats actually a real sound eq. its built into obs which is great because you cant do that in premiere pro. they have an eq but its just a random video loop, it doesnt actually sync to your mic. you can do it in after effects but it take a while to process. it would be crazy to do on a 15-20 min video. so just having it in obs to do it live as im recording takes away all the work of having to do it in the editing. really cool

thanks, i am still working on the lighting and color grading so the next couple vids are gonna be even better!

Thanks man, the video covering the features and settings is coming in about 2 hrs!