forgot to mention im using the online workflow via comfy

so it should be plug and play, but didnt work for me at all :(

hi! which error are you getting? are you running it on runcomfy? if there's an error I can flag it to them and have them fix it.

@@risunobushi_ai I'm so sorry i launched the wrong app, i launched the portrait IC by mad :)

@@risunobushi_ai I'm sorry i was opening the wrong version, i used the portrait IC one, my bad :)

0.30 USD for a LoRA (on a A40, including pod setup times) doesn't seem much compared to Replicate and CivitAI's pricing, which are arguably still rather cheap for professionals

I'm definitely not an expert on AnimateDiff, there might be better LoRAs out there, but at the stage I'm at with these workflows right now I'm focusing on finding the limitations and the areas of applicability of the tech, so I haven't shopped around for better solutions yet. I'm also very lacking in knowledge about how AnimateDiff came to be, and how it's evolving, and that's something I'll need to look into sometime in the future. Otherwise I'll just keep kicking the can down the road and settle for other people's solutions to issues I don't really have a proper understanding of.

I might be able to read minds, beware

are they questionable in the sense that they're the first tests out of a bit of tech that came out a couple of weeks ago, raw and unpolished? yes are they questionable in the sense that they showcase the complete lack of possibilities in pursuing motion graphics animations with these nodes? I don't think so. is it better to do motion graphics, right now, with anything but stable diffusion? probably, yes. my reasoning in pursuing this is that it's a neat piece of tech that makes it possible to do something that a) has a commercial application, and b) was not possible with SD before now. whether or not it's actually something that will eventually be commercially viable is a completely different matter. on the first video I published about product relighting, there was a ton of people questioning color shifts and detail loss. over the course of the following months, we fixed that, little by little. I'm trying to do the same thing here for motion graphics.

@@risunobushi_ai Thank you for clarifying. I'm not against it, it just seems like a lot of work from the side of the person that hasn't learned Comfy yet. :) I've fiddled with it, but what you guys do is way out there. Like might as well learn Houdini if I'm trying this niche :D It's nice you get to experiment though.

oh definitely! right now, learning houdini would be the smartest choice for this kind of stuff, but then again I do like experimenting with new, unconventional ways of doing the same things that other softwares do better - you never know what could happen down the line. but yeah, I was actually on the verge of cataloguing this as part of my Stable Diffusion Experimental series, but it's the existing commercial application of the motion graphics sector that pushed it over the edge, even if it's really raw tech right now.

thanks! I don't have a Patreon, mostly because I don't have the time to do more in-depth, subscriber only content (I do this for a living, and I'm either bound by NDA on more in depth stuff, or I straight up don't have the time to create more content), and because I try to share everything I can share openly, without subscriptions. I do have a ko-fi page in the video's descriptions, but that's just for donations, there's no content locked behind it.

I forgot that you had a ko-fi page. I’ll sign up over there then. Yeah, I kinda guessed you were a bit too busy for Patreon. Anyway, thanks for sharing openly and for consistently bringing some clarity to these topics.

hi! this is a known issue, the Image Level Adjustment was updated and it broke the range. I haven't had the time to fix this yet because of my job, I'll try to do it as soon as I have the time to. Unfortunately I can't maintain all my old workflows on a daily schedule.

100% 20/20 [01:26<00:00, 4.32s/it] Unloading models for lowram load. 1 models unloaded. Loading 1 new model loaded completely 0.0 319.11416244506836 True Processing image 1/1 with shape: (1448, 1086, 3) Prompt executed in 93.81 seconds thats the end. Its ksampler in regeneration group. Whats the problem?

What’s your hardware setup? What OS are you running?

@@risunobushi_ai google colab

ah, that might be why I'm having trouble helping you specifically. I can only test on Linux and Windows, I do not test for macOS (no CUDA libraries), Pinokio and Google Colab (different environments from what I'm used to).

Hi! That’s usually the job of the AnimateDiff Motion Lora. If you’re getting choppy frames, it might not have been triggered properly. Try rerunning the choppy animation with a different seed with anything else at the same values. Sometimes it happens to me too on first generations.

@@risunobushi_ai Thanks :), that was the problem. It wasn't triggering correctly. I had to skip the IP adapter group, even though it was bypassed alrdy ?¿.

yeah, you should definitely place the controlnet models in the controlnet folder. you can download them directly via the manager.