There’s two ways, you can do a clean install like the video, and copy your models over. Or edit your webui file with the git pull command at the top. I’ll auto update when you run it. I suggest doing a clean install so there’s no conflicting files. But it’s not completely necessary.
@@sebastiantorresvfx I ended up doing the git pull command and updating. Even so, I'm unimpressed with the performance so far. And I've got a 3090 with 24GB of vram. I'm trying different command line arguments but I'm just not seeing the performance boost I was hoping for. Going to investigate invokeai and comfy.
That is a very good question, might need more information? Like what kind of machine are you running it on? And what settings are you using when trying to generate with it?
Awesome! I'll try this model out soon. I have been using LoRa(s) and have gained much more knowledge into knowing what negative/positive prompts are. Though it took many trial-and-error and a lot of hellscapes, it also took a lot of tries until I eventually managed to get some of it fathomable/right. Maybe an upscale tutorial next? [PS, Thank you for providing such detailed and creative tutorials!]
Love upscaling and inpainting, though I tend to get carried away some times and end up with gigantic images 😂 Thank you for you kind words, I appreciate it 😊
Hey Seb, I watched your Stable Diffusion setup video, and I successfully installed Stable Diffusion on my system. However, when I try to generate a simple image like "A group of friends having fun at a party," it uses the same face for all the people, and it generates a whole image with 20-30 people in it, even though I tried to limit the number of people in the command. Also, it duplicates objects and people. Is there any setting to stop this?
@@sebastiantorresvfx Model - Realistic Vision V5.1 VAE and inpainting version (I just followed your last setup video) Prompt - A group of friends having fun at a party
I went with “people at a party, fun, good times” And the results were a few people in the foreground and a whole bunch in the distance. I like to generate a whole bunch and pick the best option but sometimes with the 1.5 models and really the SDLX also if I have to be honest you’ll end up rendering the same faces on a couple people of the image. Easy fix, switch over to inpainting model and tab; paint over the face you wish to change, tick masked only, then set the prompt to a random name from the random name generator. It’ll give you a different face. Do this for each of the repeat offenders. Hope that helps? Also try the juggernaut or Juggernaut XL models; the results are great with those two models.
fatal: Not A Git Repository apperd when I tried installing it.
My version still says 1.3.2.....how do I actually update to 1.6?
There’s two ways, you can do a clean install like the video, and copy your models over. Or edit your webui file with the git pull command at the top. I’ll auto update when you run it.
I suggest doing a clean install so there’s no conflicting files. But it’s not completely necessary.
@@sebastiantorresvfx I ended up doing the git pull command and updating. Even so, I'm unimpressed with the performance so far. And I've got a 3090 with 24GB of vram. I'm trying different command line arguments but I'm just not seeing the performance boost I was hoping for. Going to investigate invokeai and comfy.
Why does SDXL make my computer stop working? Why?
That is a very good question, might need more information? Like what kind of machine are you running it on? And what settings are you using when trying to generate with it?
Awesome! I'll try this model out soon. I have been using LoRa(s) and have gained much more knowledge into knowing what negative/positive prompts are. Though it took many trial-and-error and a lot of hellscapes, it also took a lot of tries until I eventually managed to get some of it fathomable/right. Maybe an upscale tutorial next?
[PS, Thank you for providing such detailed and creative tutorials!]
Love upscaling and inpainting, though I tend to get carried away some times and end up with gigantic images 😂
Thank you for you kind words, I appreciate it 😊
Seb, una pregunta: Si instalo XL, se borran los modelos que sirven para 1.5? o Quedan inservibles?
No, XL es solo un modelo. Puedes tener todo junto y elegir cuál quieres usar. Automatic1111 es igual para ambos sistemas.
@@sebastiantorresvfx Gracias, bro!
Hey Seb, I watched your Stable Diffusion setup video, and I successfully installed Stable Diffusion on my system.
However, when I try to generate a simple image like "A group of friends having fun at a party," it uses the same face for all the people, and it generates a whole image with 20-30 people in it, even though I tried to limit the number of people in the command.
Also, it duplicates objects and people. Is there any setting to stop this?
Interesting; can I ask what model you were using and the prompt?
@@sebastiantorresvfx Model - Realistic Vision V5.1 VAE and inpainting version (I just followed your last setup video)
Prompt - A group of friends having fun at a party
I went with “people at a party, fun, good times”
And the results were a few people in the foreground and a whole bunch in the distance. I like to generate a whole bunch and pick the best option but sometimes with the 1.5 models and really the SDLX also if I have to be honest you’ll end up rendering the same faces on a couple people of the image. Easy fix, switch over to inpainting model and tab; paint over the face you wish to change, tick masked only, then set the prompt to a random name from the random name generator. It’ll give you a different face. Do this for each of the repeat offenders.
Hope that helps? Also try the juggernaut or Juggernaut XL models; the results are great with those two models.
Why the hell did they change Stable diffusion?
Umm… check the date this was a long time ago
Oh I know I'm just angry they changed something that worked perfectly.@@sebastiantorresvfx
thank you
You’re welcome 😁
delete your venv folder as well, so a new one will be created with latest stuff
Good call!
thank you sir, you saved my life. had no idea why couldnt launch 1.6