This thing is EXPERIMENTAL.
It 100% will try to take over your style and character.
Please use caution. Also, weight it down to .2-.4
Mixes well with charTurnerV2
and is being released EARLY and BEFORE IT'S READY because a few prompts in charTurnerV2 mention it.
I know it has problems. I KNOW IT DOES. It's a beta. :D
Use at your own risk. :3
- mouse
Description
beta, use at your own risk.
Small grids: lora strength 0-1 over three seeds in different models.
Large grids: Lora strength 0-1 vertical, CharTurnerv2 strength 0-1 horizontal, in different models.
(uploading error, will upload soon) Clip Skip: I use both 1 and 2. There's a protogen grid showing both clip 1 and clip 2
FAQ
Comments (30)
Thanks for sharing your experiments. I think it's good for the community to see works in progress.
Sorry for asking a dumb question, but what "take over style" means? Style is already set with the prompt and model you use. Or you mean the img2img when you put already generated character and what to see it from other sides?
the Lora has a 'look' in addition to the pose, and right now, it's hard to get the turn pose without the 'style'. For instance, if you use it at 100% strength, you'll get images that are very same-samey, as the lora has 'averaged' the styles of all the training images and decided it goes with the turn around pose. For instance, it likes brown shorts and hoodies, boots and orange shirts.
Oh, so you mean it greatly affects the generation with it's own style that was taken from trained images?
@drakmour Yes, though it doesn't match any of the training images. it's like... a mix of them. It's not bad, I just don't want to impose an art style or body type on characters, I just want the pose. That's why it's an experiment, it's a very naughty lora right now.
@mousewrites we all hope that you will make it better! :-D
Could you add some example generations so we can see what it does?
It's so raw, i really only intend it to be used with the CharTurnerV2 embed at a very low strength, so I didn't want to put up a bunch of "examples" that are like, .1 of this lora, and the rest that Embed, but I will put a few up with an explanation.
added some. :)
"EXPERIMENTAL! It 100% will try to take over your style and character. Please use caution."
Why is this worded like it's some kind of virus?
because I don't like how it makes everybody a military grunt and I wanted to warn people. :D
It's definitely worded like it might accidentally murder your gen... but (virus wise) it's actually "safer" than the embed version since it's in the safetensor format. ;)
it works ! with a bit of fiddling and picking good dimensions
YAY! I'm glad it's working for you. I'll work on getting it better. :D
Is very cool! Does this not work with Img2Img? I can't seem to get it to work...
It works best with inpainting models. On the embed version of this there's a mini tutorial on how to do it. https://civitai.com/gallery/79561?modelId=3036&modelVersionId=8387&infinite=false&returnUrl=%2Fmodels%2F3036%2Fcharturner-character-turnaround-helper-for-15-and-21
Excellent work! Is it really possible to generate more than 3-7 poses by rotating the character? Thanks
between this and control net, yes. If you have particular poses you want, control net can help. CharTurner keeps people in the same outfit, so you can use a 'mulipose' photo shoot and get mutiple poses.
Having trouble getting results that don't slaughter the picture, I tried recommended strengths, several models. My SD can't do TI's so I'd love to see continued development of this Lora.
sorry, where should i put?
Lora go in the models/LORA folder, embeds (the other file) goes into models/embeds
THANKS
@mousewrites
Hello, excuse me, I put CharTurner-Character Turnaround helper for 1.5 AND 2.1! This is in [embeddings], but can you tell me how to confirm that it's in use, or how to activate it, can you tell me
In addition, my creation is about ducks, but all of them are human beings. Is this lora suitable for animals
There is an embedding version of this, as well as a lora version. You commented on the Lora version. Either one works, just put the lora in the lora folder, or the embed in the embedding folder. Make sure you get the version for the type of models you are using (1.5 or 2.1).
It should be able to do duck people if they are bipedal; there's some images in the embedding folder of badger people. :D
Thank you for your answer. I have taken a look at my model Disney Pixar Cartoon Type A, which displays [Base Model SD 1.5]. As I am still learning, it may be due to an error I made somewhere. Thank you again for your answer
:)
@SharedObjects Yup, sounds like the 1.5 version is the one you want. You can use either the lora or the embed, or both at once (but that doesn't usually work great. XD)
As a note for future users, 1024x512 worked best in my testing, feel free to leave comments with your best resolutions :)
i don't know what happens but My stable diffusion just doesn't recognize this file as a LORA.
tried to change the webui some few times and still same.
With a resolution of 768x768 I can get a front view and a side view?? I can also help myself by using the Pose controlnet.
Should be able to, but openpose helps nail things down, of course
Details
Files
charTurnBetaLora.safetensors
Mirrors
charturnerbetaLora_charturnbetalora.safetensors
charTurnBetaLora.safetensors
charturnerbetaLora_charturnbetalora.safetensors
charturnerbeta.safetensors
charturnbetalora.safetensors
charturnerbetaLora_charturnbetalora.safetensors
9.safetensors
charTurnBetaLora.safetensors
model.safetensors
charturnbetalora.safetensors
charTurnBetaLora.safetensors
charTurnBetaLora.safetensors
charturnerbetaLora_charturnbetalora.safetensors
charturnerbetaLora_charturnbetalora.safetensors
charturnerbetaLora_charturnbetalora.safetensors
charTurnBetaLora.safetensors
charTurnBetaLora.safetensors
charTurnBetaLora.safetensors
charTurnBetaLora.safetensors
charturnerbetaLora_charturnbetalora.safetensors
charTurnBetaLora.safetensors
charTurnBetaLora.safetensors
charturnbetalora.safetensors
charTurnBetaLora.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.









