All images have WF.
STR 0.62 on conditional and 0.25 on unconditional. This thing is Incredible, it just need the correct prompting, use a good workflow and a good prompt you will be very impressed.
Description
First Version
FAQ
Comments (15)
Does this great LORA work with the ostris/ideogram_4_turbotime_lora? I like IG4 but it's so slow as you need CFG 7 by default even on the turbo 12 step version. If you have time as your probably very busy can you please do a LORA for Boogu-Image Turbo as it's a great open source model with the Apache License 2.0. Thank you
Great LoRA! 0.62 for cond tends to give you Caucasian women if you mention "fair skin" or "pale skin". It works better with lower weights like 0.20-0.30 when you are making Asian chicks, imo.
couples of x missing from mystic? is that a mistake? there is new ideogram model type available in filter. no need other type.
'here is new ideogram model type available in filter. no need other type.'
If I remember correctly, due to licensing restrictions, Civitai stated that they would remove all NSFW Ideogram 4 LoRAs. Categorizing them under the 'other' section might be a way to stay under the radar.
read the license
are you able to provide a list of concepts this was trained on?
they kinda have
woah 😳
Who was influenced by Turbo Lora, resulting in characters with large heads and small bodies?
This is crazy good for a V1! Helps a lot with variety of faces and body types. Excited to see how you can improve it. Thanks for sharing your work!
Looks like fun! How do you get your prompts, lm studio or ollama or something similar? I'm guessing chatgpt is off limits lol
Local qwen 3.6 27b, cloud deepseek v4 api key free
After 3 very frustrating days trying to get high quality pics with ideogram i saw your pics and thought, why his pics look so much better than mine. I just added your lora and BAM, the ultra shitty lighting from ideogram instantly got fixed (seriously why would they release something so dark). You are goated bro thanks a lot.