Gibi is a YouTuber who is mainly known for making ASMR, which is a genre of videos about sounds and e-girls. She also streams on twitch and does cosplay.
The embed has been trained on SD 1.5 and works well on most 1.5 based models. Seems to tend the variety of clothing on some models, but increasing the token weight or using inpainting helps out.
If you enjoy my creations and want to support me or have a face you'd like to commission, consider supporting me on Ko-Fi or Buy Me A Coffee :)
Description
FAQ
Comments (14)
Didn't look like her though
"ASMR, which is a genre of videos about sounds and e-girls"
How I can tell you've never seriously watched/listened to ASMR, lol. Nice embedding though.
I like Ephemeral Rift a lot because of his infinite creativity, dedication and passion, and most asmr pre-e-girl wave. I've watched plenty, I'm just goofing around :D
I've got a dataset for Frivi that needs captioning but I'll train and release an embed of her soon.
@SDKoh 👀 still in the works?
@furAIenjoyer I kinda forgot about this one due to personal circumstances, but I've put it back on my list ^^
@SDKoh thanks! :D looking forward to it!!
@furAIenjoyer I'm very sorry to inform you that the first training run yielded atrocious results. I'm not sure why, but SD overtrains her face very easily. This basically means that so far I've only managed to train caricature versions of Frivi. I have a few ideas as to how to prevent this, but there's no guarantee it will work.
@SDKoh aw damn, I hope you can get it working!! maybe a Lora would work if you're able to train those?
@furAIenjoyer It's worth considering a Lora if an embedding doesn't work out. I have a preference for embeddings though. I think I'll have to recreate the dataset and take it from there.
can't run the model on a 7900 XT, i'm getting the "SystemExit: Attempting to deserialize object on a CUDA device but torch.cuda.is_available() is False. If you are running on a CPU-only machine, please use torch.load with map_location=torch.device('cpu') to map your storages to the CPU." error :/ i'm using shark sd and most civitai models works well, a fix would be appreciated !
I have never seen this error message before, I'm honestly not even sure if this is on my end given that it alerts you to torch/cuda, which are local. It's an embed trained like any other, nothing special about it. I'll get in touch with some fellow creators and see what they have to say on the error, though. Maybe there is something on my end I can do after all
Just to clarify, are you only using your CPU to run Stable Diffusion, rather than the GPU?
Hi, I did some troubleshooting, mostly on StackOverflow, and there does seem to be a compatability issue for models trained purely on GPU's to run on pure-CPU setups. However, reports on this are mixed as this does not necessarily seem the case. Regardless, I find it hard to find a solution to this problem.
Could you confirm whether or not you are running SD purely through the CPU? Have you tried to apply any of the common answers on StackOverflow in regard to this error message? I can't link posts from StackOverflow due to it not being a allowed URL, but if you got StackOverflow and add '/a/68992197' after '.com', it should lead to a thread about the issue.
@SDKoh Hey, thanks for your response ! I'm running stable diffusion on my GPU. I'll check on StackOverflow to see if i can solve the problem, thanks for the tip :)
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.