CHAPTERS
0:00 Introduction
0:16 Prompt editing
3:31 Prompt alternating
In this Stable diffusion tutorial I’ll talk about advanced prompt editing and the possibilities of morphing prompts, as well as showing a hidden feature not yet revealed in the wiki.
Ultimate stable diffusion tutorial:
https://youtu.be/DHaL56P6f5M
I love your Ai videos above all creators. ā¤
Is this a new model? I mean do you have to download something to get it to work
thank you …again
what website is that one for stable diffusion??
is it locally in your computer?
amazing content, fan from brazil!
Thanks friend!
2:32 omg please tell me you've saved that image haha š
Thanks! š
Why, out of curiosity, did you use Heun?
Can you explain where this is available?
Great tips thanks man!
Do you know how to train your own faces on S.D?
How did you make it show a preview image during processing??
Seb you have a talent for this youtube stuff! Good job!
is that live preview new? i dont have that
bro can you make a video on how to update the automatic1111 gui, i have the outdated version , no idea how to update them
Thanks, helps seeing it happen visually like that.
great video! wondering how is this different from prompt weighting? is the difference that the prompt changes per-step as opposed to combining both into percentages for all steps? seems like it has a similar outcome. Thanks!
Some reason to use HEUN Sampler?
Too bad stable diffusion doesn't work with AMD GPUs
Thanks!
1:29 You said that [llama:tiger:25] would generate a llama for the first 25 steps, then a tiger for the next 25, however in the preview we can clearly see that this isn't what's happening. After 25%, it first attempts to generate a tiger, then at 50% switches back to llama, then finally at 75% presumably finishes with tiger.
Perhaps it is switching every 25% rather than every 25 steps, although that doesn't explain why 0.5 gives the same result.
You kind of described that after 50% it merges both promts together, which might be a possibility but I would expect the Promt Editing to give you full control.
Is there anywhere we can read up on the additional prompt syntax?
dope
how is your webUI showing a live preview of what it's generating? Mine just loads then gives me a photo. I'm running locally on a 3060 12 gb
That's really neat I wonder if you could use the prompt Matrix to make an animation
so usefull man! its game changing !! thanks a lot
Thanks for this. Excellent
Very very informative! Greate video. Thanks!!
I always thought that a "liger" is a crossbreed between a lion and a tiger. Who thougt that it would be a llama and tiger!
I love that this feature exists and at the same time I have no idea what to use it on š¤£
Great tutorial, cheers! Does anyone know if there is a way to output the steps as individual .png files? Would be great for morphing animations! And do you know if deforum has any similar tricks to its prompts? Thanks
No prob"llama"… "Alpaca" my bags and go… urgh…. URGH. But, well done š
merci
very cool!
God bless Automatic1111
Can you make a video about merging models? Do you know if it is possible using colab?
Does this only work on heun?
ššš
How did you make your SD output individual steps as it generates the image?
Alpaca your bags in a second #2!!
cotch in a tumble dryerš
The ending was funny xD
I appreciate you! Amazing tutorial. Fun stuff!
how did you get your SD to show a preview of what its doing as before the image is done?
The whole video was great and the joke at the end just was the icing on the cake, great job š
no probllama.. alpaca my bags.. t-shirts!!! š
ā”ļøš
Your image is updated in real time while it is generated. What option is Settings allows it? By default you just look at empty Preview area and at the end you get a final image
How do u view the steps….. i dont see the steps just the final result
It works. Thanks!
Very cool
Woo, neat! I'm wondering if it works in Stable Diffusion UI. I'll try it.
Hello sir, I made a dreambooth model of myself right, the only problem is that I would want to also train a style for myself, how do I merge the two? I can't use 2 models at the same time
bob ross is so happy
Which SD GUI is that? It has so many settings lol. Curious about trying it out.
no problama… I see what u did there…
Maybe all these AI content creators should start added a credit list in the info section for all the artists they use in their prompts. Maybe provide a link to their websites to give them exposure as well. Just a suggestion.
Excellent
Your content is great! Keep the hard work man
"…Put a stool in there and it will look real!!!" hahah the commentary has me dead
Wow, that's cool. Is there any way to do this with 2 images? I really could do with being able to merge 2 images into a third result.
coooool
hello , Every time I try to run webui-user it always shows that it is updating then after quite sometime it gives me a error message about updating then ask to press any key to continue, but torch and torchvision not installed, can you tell me what is the issue of this installation process?
this is called morphing and we can do it at least 5 years ago
AI has a lot sense of humor
I like how you used Zabrocki bc it basically looks like *he who shall not be named*, but without the drama. š
I am having a very strange experience with this. I am using [Apple:Tree:0.66] 100 steps and at steps 0-34 it is rendering a tree and 34-100 its an apple.
I guess I could just flip the words, but I was under the impression that the input had positional priority…?
hi , can you share the link of this? is very interesting
Btw the vertical bar you can type with Alt+124. Your welcome š
I assume this has to be done on a locally installed SD?
What are the advantages of telling the AI to alternate between prompts vs. just including both terms directly in the prompt itself?
Great video. Iām wondering if you know what to prompt to make a 2D plan floor from a roof image. So just the outline of the roof from top view down .. that way no side walls would be shown. Iām trying to create a 3D model of a house I really like and it would help if I can create the floor plan to help me build the walls in blender. Or even sketch fab.
Wow!! you showed me a new one i have not seen yet. thank you for sharing this!
I think that the Prompt Alternating method can sometimes be good for for making an optical illusion such as the famous of albert Einstein Marilyn Monroe illusion
for example , you can see the picture of panda here
https://youtu.be/WwX_7HIJBTM?t=246
https://youtu.be/tB5-JahAXfc?t=7
holy crap! look at the panda rose and squint your eyes till its blurry…. it straight up turns into a panda xD
0:24 Your move Ben Shapiro
Hi, I've tried this, but the transformations only work when I use the braces "{}" instead of brackets "[]"what is going on? They are constantly changing things…is this again, one of them?
can we change prompt on the fly ?
AI just, "this doesn't make any sense" and then just paints the tree on a box.
Wanted to make a flame-themed brooch for an ember witch I generated so I tried the prompt editing trick for the first time. Worked amazingly. Thanks!
what a waste of time
How do you get the preview of your SD processing? š
Does this feature work in the colab or is it something that needs to be built in manually?
sadly, doesn't work on AMD GPU's. š I have found an AMD-friendly non-GUI Stable Diffusion but there's no UI. :/
No prob-llama bro š
This is AWESOME – combine with video-frames and you have total magic. No Toto we're not in MidJourney anymore.. real AI.
Wonder if you can apply to Dream Booth too?
Your puns made me subscribe. I hate you for it. Thank you. [appreciation,anger: 50]
My results never look as good as yours. I am running my SD using the low end PC code line. How much of a difference does that make?
I've never seen the computation happening live, only the end result… what setting is that?
your channel is the most useful one in regards to learning about AI-Stuff, which I found so far. So thanks alot, mate!
Very cool thanks!
Skull_rose_llama would be a great addition to any rock bar's wall.
I use alternating prompts to mix a lot of faces without going over the token limit. "a female [elf] ranger, played by [all|your|favorite|actresses]".
omfg, every furry fetish dream come true, we can do human:cats ratio
Incredibly engaging, thank you for this. I don't even do such advanced things but you make it so much fun!
"Now let's end this with something more positive" "He is with us no more, here are his llama friends mourning his passing. Goodbye, dear llama." Truly more positive.
Thank you so much! This is definitely helpful.