@Dr. Furkan Gözükara Do you also have side by side comparisons for unet only + now that we have clip
@Dr. Furkan Gözükara Do you also have side by side comparisons for unet only + now that we have clip_l training aswell
Fyes i have
Fbut i didnt perfected clip l yet
Flet me update patreon post for early experiment
SSounds good
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
SAwesome will have a look
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=40)
DJust out of interest Doc when generating are the majority of the images holding the same likeness in quality? do u get many images do you get that have some deformity or have no fidelity?
![Dazzastrous [4090]](https://cdn.discordapp.com/avatars/871389071686135838/c4a5f4d81d9eeb18d856295d692e2c07.webp?size=16)
Falmost all perfect
Fwith fux
F
Fi am mass testing prompts atm not like generating a lot of same prompt
Dhehe mars
Fbut looks like we really should add different angles
Flike looking down
Fup
Fleft right
Ffor such prompts
Fit works really good
Dsmiling , angry, confused
UHey! Do you know if there is a way to specify image color palette for Stable Diffusion or Flux, without using Controlnet extensions?
Finclude in train dataset and should work perfect
Fnot that i know
Fi am testing torch preview nightly
Fon windows to see flux speed
F
F
Fwow torch 2.5 brought speed up to
Fwindows
DIf you promt for cinematic colors, cinematic lighting you definitely get a different color palette, may be you can specify like teal and orange, or other well known color palettes
UI'm mostly looking for illustration color palettes actually. I'm trying using ChatGPT to give me color palettes with most common names of the color tones.
Dgreat news, how it compares with Linux speeds?
Falmost same i think
Fi will udpate my installer and add it
FHello, do you know a way to upload files to massed compute like runpodctl? Currently I upload files to a drive that I retrieve but it takes a long time even for small files. Thanks
Dcan you do a fix .bat to not have to reinstall kohya completely?
Ddid you updated manually with a pip install from the venv?
�You just gave me an idea to see if Flux can understand hex codes from the prompt:
�Looks like it does have some kinda effect lol
Flet me upload
Fdropped from 8 second to 5.8
�Create beautiful FLUX images in milliseconds with FastFLUX. Free, fast, and no sign-up required. Image generation powered by Runware.
UWow. I tried on SD models it did not work but haven't tried Flux yet. Good job. I think it's doing good but can you try more?
�Coolors.co
Generate or browse beautiful color combinations for your designs.
UWe may even create wildcards and/or lora's with this
�Hmm, I think perhaps using IP Adapter with just a color palette may work better tho. Haven't tested it out yet
UYes but I want an easier way. Maybe it could be color palette presets. I could define color palette with common names + hex codes for each color style
UIn illustrations, it's really important to define key color palettes to have a unique and constant art style
FPatreon
Get more from SECourses: Tutorials, Guides, Resources, Training, FLUX, MidJourney, Voice Clone, TTS, ChatGPT, GPT, LLM, Scripts by Furkan Gözü on Patreon
F28 August 2024 Update
340 prompts txt added to the testing prompts list
Install_Torch_2_5_Dev_Huge_Speed_Up.bat file added
This will install nightly development branch of Torch
It brings huge Speed up on Windows and you can get closer speeds to do training in Linux
I haven't tested its quality yet so it is experimental
340 prompts txt added to the testing prompts list
Install_Torch_2_5_Dev_Huge_Speed_Up.bat file added
This will install nightly development branch of Torch
It brings huge Speed up on Windows and you can get closer speeds to do training in Linux
I haven't tested its quality yet so it is experimental