Me: "Read the source code, it's not hard" Also, Me: *visibly struggling to read the source code* The formatting used a torch function instead of just raw math, so I had trouble finding it: tw = torch.mul(t1, conditioning_to_strength) + torch.mul(t0, (1.0 - conditioning_to_strength)) Which is, effectively, the same formula that I wrote, so everything is fine.
my audio interface died, so I was unable to use my studio microphone. I had to use a C922 webcam's microphone to record the audio. if you don't know about that camera, it's actually THE recommended one for the lowest possible audio quality.
It seems that putting both prompts in one box separated by a comma yields most of the time something similar to using concat but it sometimes pick one or the other for instance you sometimes get a bowl of apples only or a girl with white hair alone. Concat is 100% consistent and it never separates the concepts. Thanks for the explanation.
Do you ever script in python? Cool stuff I haven't touched SD since May 2023 I'm no artist but I wonder if you'd like it more than visually scripting with ComfyUI, but I expect its nice to have the flexibility and visuals. You might be able to make an attached program for ComfyUI in VSCode for another art software. 3D, or adobe ps I'd absolutely use ComfyUI for something like a cool dynamic threshold controller and weight cutoffs between models.
ik you have a video on this but cud you explain how to upscale images correctly i have tryed you method but it takes 67s/it for me to do so i need some way of upscaling and still adding more detail without having to wait 20 minutes for one upscale INFO: I'm on amd win11 currently trying every way of upscaling (which i know of) with different settings and comparing them all untill i find something good that dosnt take ageees to upscale. if you want to help me out of this hole you can tell me a good method you know or use. thanks :)
@@poisenbery I have a 6650xt it’s like a 3060-70 and it has 8 gb vram is it really on the card that it takes 70s/it to upscale ? Bcs if I do my usual stuff I get to like 1.3s/it or up to 3s/it at max…
Me: "Read the source code, it's not hard"
Also, Me: *visibly struggling to read the source code*
The formatting used a torch function instead of just raw math, so I had trouble finding it:
tw = torch.mul(t1, conditioning_to_strength) + torch.mul(t0, (1.0 - conditioning_to_strength))
Which is, effectively, the same formula that I wrote, so everything is fine.
I’ve never been more happy to listen to the world’s worst audio 😂. Thanks for the information! I really like your way of explaining AI
Appreciate the information. Not sure why you decided to deliver it over an airplane's loud speaker, but hey, you do you.
my audio interface died, so I was unable to use my studio microphone.
I had to use a C922 webcam's microphone to record the audio.
if you don't know about that camera, it's actually THE recommended one for the lowest possible audio quality.
That's actually a great video, I never understood the difference and when to use them, now I feel like I understand it.
Great to see that you still have internet from your spaceship ! 😅
with the help of a combine you can mix day and night, or summer and winter, that's what suits it
Nah bro i love this channel, hes really funny and entertaining😂😂😂😂😭😭😭😭I just hope he doesn't get a better microphone this is him now lol
It seems that putting both prompts in one box separated by a comma yields most of the time something similar to using concat but it sometimes pick one or the other for instance you sometimes get a bowl of apples only or a girl with white hair alone. Concat is 100% consistent and it never separates the concepts. Thanks for the explanation.
What about in negative prompts?
And what if we're using prompt masking?
Amazing content! Thank you so much for this.
Didn't know I was on the subway... No I don't want extra frise or to upgrade my meal.... Bruh the mic XD. My mind grows but my ears shrink.
Amazing! Thank you!
Great explanation but you forgot one thing, you need a fan blowing directly into your mic the entire time. 9/10
Do you ever script in python? Cool stuff I haven't touched SD since May 2023
I'm no artist but I wonder if you'd like it more than visually scripting with ComfyUI, but I expect its nice to have the flexibility and visuals. You might be able to make an attached program for ComfyUI in VSCode for another art software. 3D, or adobe ps
I'd absolutely use ComfyUI for something like a cool dynamic threshold controller and weight cutoffs between models.
I love your videos poisenbery, but man for the love of god upgrade your microphone. I beg of you 🤣
The audio was bad? I was too busy getting knowledged.
ik you have a video on this but cud you explain how to upscale images correctly i have tryed you method but it takes 67s/it for me to do so i need some way of upscaling and still adding more detail without having to wait 20 minutes for one upscale INFO: I'm on amd win11 currently trying every way of upscaling (which i know of) with different settings and comparing them all untill i find something good that dosnt take ageees to upscale. if you want to help me out of this hole you can tell me a good method you know or use. thanks :)
You need a better GPU
@@poisenbery I have a 6650xt it’s like a 3060-70 and it has 8 gb vram is it really on the card that it takes 70s/it to upscale ? Bcs if I do my usual stuff I get to like 1.3s/it or up to 3s/it at max…
The audio is bad but the video's worth watching
For a second I thought you'd start reading out the Wikipedia article about CBT.