0:00 Intro 0:15 Where Is The Lora Section Located In Forge 0:59 The Best Place To Find And Download Loras 2:11 Important Information To Take Note Of When Downloading Loras 3:14 You Can Also Find The Prompts Used For Generating The Images 3:43 How To Install The Loras With Forge And Flux 4:59 Editing Some Text Settings With Our Lora In Forge 6:03 VERY IMPORTANT INFO To Make Sure The Loras Actually Work In Forge 7:20 Inserting And Using The Lora In Forge With Flux 8:44 You Can Use 2 Loras At The Same Time 10:00 Outro - Thank You For Watching
@@petertremblay3725 yes you can do this, I use this method for face swapping. So if I trained a Lora on someone’s likeness, I could just inpaint their face and even their hairstyle onto another image, works like a charm 👍🏽👍🏽
my lora files doesnt show.... i dont know why...that bug....on forge ui, when i add sdxl fordel with lora inside in lora folder it only show sdxl folder,.
for some reason it is not working for me. No matter which Lora I use, the results are always the same with and withouth the Lora, I'm using fp8 checkpoint, yes I haev the automatic (fp16 LoRA) option enabled
I've had a bug for a while now that I can't add LORA by clicking on it. It does work if I type it manually, but it's annoying to do so. Does anyone know why this happens?
Hey, so I've ben using CiviAI to generate images using PonyXL but when I try doing the same using A1111 or Forge I cannot for the life of me get them to even look remotely similar, it's like I have some setting turned on/off incorrectly and the result is total poop. I've 100000% made sure to have the exact same prompt and negative prompt, Loras (and their strengths), the model, image size, clip skip, etc. Even with the same seed (which I know is different for my machine and CivitAI - but should just make slight variations) it's a no go. My assumption is that it's because I'm using LoRAs (style LoRAs) that don't have a trigger word/activation text and then when I try putting them into the prompt text it messes it up. But idk how to do it correctly then. Because on CivitAI you can add the LoRA to the generation window and it just works, and you slide the strength up and down. Any advice?
@@pochabhai Here you go, this is how I setup and installed Forge and FLUX locally on my pc, by default the interface will look exactly the same as you see it in this video. ua-cam.com/video/yGotOmz6LpQ/v-deo.html
Thanks. I was using 1.5 for a while and switching to flux now. It seems all the old Loras are not working. Is that correct or is there something to do? thanks
You can't use stable diffusion 1.5 loras with Flux, they won't work. You will need to download flux loras, if you go to civit ai there are tons of them to use with flux.
Go to this video and scroll to 7:28 if you want to see which flux model would work best with your hardware. ua-cam.com/video/yGotOmz6LpQ/v-deo.html Requirements GPU: Ideally you want A high-performance GPU as it's highly recommended to handle the image generation process. FLUX is optimized for Nvidia GPUS For optimal performance: NVIDIA GPUs with CUDA support, such as the RTX 30 series or above, are typically required. You can still use A GPU with 6GB of VRAM, such as an NVIDIA GTX 1660/RTX 2060. It can still run FLUX but might require lower batch sizes and optimizations like mixed precision and will obviously be much slower but it's possible. RAM: 12-16 GB of system memory should suffice. 32GB is recommended though. Storage: 30-50 GB of free space is still recommended for model weights and outputs. My installation was 96GB in size. CPU: A modern multi-core CPU (e.g., Intel i5/i7 or AMD Ryzen 5/7) Operating System: Linux (Ubuntu), macOS, or Windows.
It's possible but requires some extra work. If you try and use two loras at the same time with different people, the loras are probably going to blend into each other and you'll get two people sharing similar facial features. Inpainting can help separate and solve this. So let's say you want to generate two different females. What I would do is generate one image using just one of the loras and maybe type a prompt like "a girl standing next to another female". When that image generates, I'll disable the first Lora and enable the second one, but now i'll do inpainting with Forge and paint a mask over one of the females faces to allow the new face to be generated. It's hard to explain in text, I would need to do a tutorial.
There are apparently extensions that can be installed that can rectify the issue of using 2 loras at once of different people but I would need to test these myself first and see if it works and record a tutorial.
0:00 Intro
0:15 Where Is The Lora Section Located In Forge
0:59 The Best Place To Find And Download Loras
2:11 Important Information To Take Note Of When Downloading Loras
3:14 You Can Also Find The Prompts Used For Generating The Images
3:43 How To Install The Loras With Forge And Flux
4:59 Editing Some Text Settings With Our Lora In Forge
6:03 VERY IMPORTANT INFO To Make Sure The Loras Actually Work In Forge
7:20 Inserting And Using The Lora In Forge With Flux
8:44 You Can Use 2 Loras At The Same Time
10:00 Outro - Thank You For Watching
you sir, just gained my sub , FOR LIFE!
You're awesome! Thank you for the support and sub!
6:30 was the important part for me, that no other tutorial mentioned and finally made it work. Thank you!
Awesome! I'm glad I could help Philipp, thank you for watching.
Really helpful tutorial-thanks for sharing!
I dont have a refreshbutton at 5.00???
That's odd, then just close forge and Pinokio and restart it.
thx bro...great tutorial!
Thank you! Appreciate that.
Can lora be apply only on specific part of the image like inpainting?
@@petertremblay3725 yes you can do this, I use this method for face swapping. So if I trained a Lora on someone’s likeness, I could just inpaint their face and even their hairstyle onto another image, works like a charm 👍🏽👍🏽
@@astrovah OK thank for replying and happy holidays to you and your family.
my lora files doesnt show.... i dont know why...that bug....on forge ui, when i add sdxl fordel with lora inside in lora folder it only show sdxl folder,.
for some reason it is not working for me. No matter which Lora I use, the results are always the same with and withouth the Lora, I'm using fp8 checkpoint, yes I haev the automatic (fp16 LoRA) option enabled
I've had a bug for a while now that I can't add LORA by clicking on it. It does work if I type it manually, but it's annoying to do so. Does anyone know why this happens?
I wish i had an answer for this, that is strange that you can't click on it. Have you maybe tried updating Forge to see if it solves the issue?
Hey, so I've ben using CiviAI to generate images using PonyXL but when I try doing the same using A1111 or Forge I cannot for the life of me get them to even look remotely similar, it's like I have some setting turned on/off incorrectly and the result is total poop. I've 100000% made sure to have the exact same prompt and negative prompt, Loras (and their strengths), the model, image size, clip skip, etc. Even with the same seed (which I know is different for my machine and CivitAI - but should just make slight variations) it's a no go.
My assumption is that it's because I'm using LoRAs (style LoRAs) that don't have a trigger word/activation text and then when I try putting them into the prompt text it messes it up. But idk how to do it correctly then. Because on CivitAI you can add the LoRA to the generation window and it just works, and you slide the strength up and down.
Any advice?
how does one get that interface where you are working on this video?
This is just the default Forge interface when you install it using Pinokio :)
@@astrovah can you please share the link to the same?
@@pochabhai Here you go, this is how I setup and installed Forge and FLUX locally on my pc, by default the interface will look exactly the same as you see it in this video. ua-cam.com/video/yGotOmz6LpQ/v-deo.html
@@astrovah thank you for your help, will surely try it out
Thanks. I was using 1.5 for a while and switching to flux now. It seems all the old Loras are not working. Is that correct or is there something to do? thanks
you need Flux specific Loras, 1.5 Loras are only for 1.5
You can't use stable diffusion 1.5 loras with Flux, they won't work. You will need to download flux loras, if you go to civit ai there are tons of them to use with flux.
Correct!
what is the minimum requirements for this installation
Go to this video and scroll to 7:28 if you want to see which flux model would work best with your hardware. ua-cam.com/video/yGotOmz6LpQ/v-deo.html
Requirements
GPU: Ideally you want A high-performance GPU as it's highly recommended to handle the image generation process.
FLUX is optimized for Nvidia GPUS For optimal performance:
NVIDIA GPUs with CUDA support, such as the RTX 30 series or above, are typically required.
You can still use A GPU with 6GB of VRAM, such as an NVIDIA GTX 1660/RTX 2060.
It can still run FLUX but might require lower batch sizes and optimizations like mixed precision and will obviously be much slower but it's possible.
RAM: 12-16 GB of system memory should suffice. 32GB is recommended though.
Storage: 30-50 GB of free space is still recommended for model weights and outputs. My installation was 96GB in size.
CPU: A modern multi-core CPU (e.g., Intel i5/i7 or AMD Ryzen 5/7)
Operating System: Linux (Ubuntu), macOS, or Windows.
How could you use, within Forge with Flux, two different LORAs for a pair of people in the same image?
It's possible but requires some extra work. If you try and use two loras at the same time with different people, the loras are probably going to blend into each other and you'll get two people sharing similar facial features. Inpainting can help separate and solve this. So let's say you want to generate two different females. What I would do is generate one image using just one of the loras and maybe type a prompt like "a girl standing next to another female". When that image generates, I'll disable the first Lora and enable the second one, but now i'll do inpainting with Forge and paint a mask over one of the females faces to allow the new face to be generated. It's hard to explain in text, I would need to do a tutorial.
There are apparently extensions that can be installed that can rectify the issue of using 2 loras at once of different people but I would need to test these myself first and see if it works and record a tutorial.