i have same issue, where did you put this command?
i have same issue, where did you put this command?
SI didn't have the command, I changed over to the development branch of the git
Bthen i am so confused where to put this command
SHello, Everyone. I'm having an issue. If someone knows, could they please help me? Thanks!
I don't exactly know what is causing this. There is no error per se. I'm trying to use the SOTA Image Captioning Scripts and Clip_Interrogator Gradio Web UI but it seems much slower than the demonstration. I also have a 3090 so I should be getting the same speeds.
I'm using
ViT-bigG-14/laion2b_s39b_b160k
blip2-2.7b
Though, this is the current ETA:
I don't exactly know what is causing this. There is no error per se. I'm trying to use the SOTA Image Captioning Scripts and Clip_Interrogator Gradio Web UI but it seems much slower than the demonstration. I also have a 3090 so I should be getting the same speeds.
I'm using
ViT-bigG-14/laion2b_s39b_b160k
blip2-2.7b
Though, this is the current ETA:
S
EHello everyone!
I had a quick question for you. I seem to remember coming across conversations here with respect to having stable diffusion model(s) that allowed stylized text (word) generation, without the use of control net. For the life of me though I can't remember what or where it was. Do you know if there is such a thing? Thanks for your help!
I had a quick question for you. I seem to remember coming across conversations here with respect to having stable diffusion model(s) that allowed stylized text (word) generation, without the use of control net. For the life of me though I can't remember what or where it was. Do you know if there is such a thing? Thanks for your help!
LIntroducing Harrlogos v2.0! Solving Stable Diffusion text generation one LoRA at a time! THIS DOES NOT REQUIRE CONTROL NET TO WORK Harrlogos now on...
LThere's two out now I think, you need to use SDXL tho
EThank you!!!
No worries. I’m fully on SDXL
No worries. I’m fully on SDXL
LI hope everyone gets to 
EWith the exception of doing sdxl training, is there something that stops you from using sdxl?
Maybe I can help?
Maybe I can help?
LI only use SDXL, all day literally
EGot ya. I interpreted your message a different way I think so thought I could help. Thanks again!
LNo problem nice to meet you!
LHarrlogos isnt perfect but its pretty cool
L
L
LThe bottom one I modified
ENice to meet you too.
These are great! Looks like I’ll be up late tonight
These are great! Looks like I’ll be up late tonight
Lme too probably, I just did a merge and I'm going to build a tensor engine for it, I pray that the SDXL Style selector works then I'll have what I want. The merge went great.
EOh cool you build as well!
I don’t build tensor engines but I am working on projects merging several AI concepts. Happy to DM and chat more if you’d like to chat more.
I don’t build tensor engines but I am working on projects merging several AI concepts. Happy to DM and chat more if you’d like to chat more.
LWe can chat here it's all good
AYouTubeAndyAI Insights
Dive into the enchanting world of Christmas with our heartwarming video that captures the essence of the holiday season. Join us on a magical journey filled with twinkling lights, joyous melodies, and the spirit of giving.
As snowflakes gently fall, families gather around crackling fires, sipping hot cocoa and sharing laughter. The streets are ...
As snowflakes gently fall, families gather around crackling fires, sipping hot cocoa and sharing laughter. The streets are ...
Fhello did you fix your error?
Fyes you are getting into shared vram
Fsomething must be wrong
Fsomething else on your computer must be using your VRAM
Fdmn this is nice lora
Fthis is great
LI used Adobe to fix the R and add the fun dice. I supplement my pictures there to put the finishing touches
E@The MisoGenius , it's almost 2am here and I've been playing with this. I will say, it is nice, but it does have limitations. If a word is more than 4-5 letters, it starts getting confused. If it's a popular phrase though like "happy new year", it'll do better. Definitely not an "end all be all" model, but it does get you started. Thanks again for pointing me to it. Here is something I generated. It did take a lot of inpainting though as it can't follow this many letters, especially if you decide to change up the spelling. But, overall a really nice way to get you going and creative.
LWhat I will say is the technology will become capable of anything but the art is limitless
EYea absolutely the art is stellar I have nothing to say towards that. It's incredibly creative.
LI'm launching my site probably tomorrow along with my socials im just wrapping up its why im up late. The project I felt like doing earlier was I trained my subject but i captioned him as a woman and I used women for Class images because I wanted to use loras for womens clothing and to make it more versatile. And I was right.
ESimilar. I'm hoping to get my site out this week.
LSweeettttt
NThank you Furkan for your help.
N
NI recently discovered Decorify (https://decorify.app/) and tested it with a living room image. Unlike my attempts with ControlNet, which maintained the furniture's structure, Decorify generated images with varied styles while preserving the room shape and rearranging the furniture in different shapes, like the diverse coffee table designs in each image. What might be their approach, and how does it differ from ControlNet?
NSo far I have tried controlNet with mlsd, depth , segmentation and canny
JSo i had a chance to go through the Zero to Hero Stable Diffusion YouTube (great video btw). Althought our versions are slightly different, I was able to identify and set all the settings the way you did. Unfortunately, when I go to train it, it's just generating black pictures. So... I might be missing something. Any thoughts?
NDear Furkan, thanks a lot for your continued support.
Fhow is your input photo?
Fhi. welcome
Fi dont suggest to generate samples during training
Fmoreover that extension is now sadly outdated
Fif you dont have gpu best is renting from runpod
Fif you cant afford runpod we have kaggle SDXL DreamBooth
Fand it says you have 6 GB VRAM
Fyou can't do training with 6 GB GPU
Fye ye sure to get scammed . bye
to get scammed . bye