a LoRA model based off Cacao from the Nekopara anime adaptation
6.1 MODEL RELEASE DATE: Never? (By me atleast)
Whoever would like to create Cacao 6.1 is happy to do so, full permission granted. Just make sure to ping me so I can see the full product myself :3
Here is the link for the zip file as I've abandoned this LoRA and LoRA creation in general:
Link
Very late on this but I thought I give updates on the why:
When I was making 4.0 Google Colab was free to use with Stable Diffusion, 4.0 became a version that is usable for the most case but only for creating Cacao, so there wasn't many variety, and generating those were painfully hard.
6.1 consisted of 287 images, literally every scene where Cacao appeared, or was in a different outfit got a screenshot, still image. I literally rewatched the entire Nekopara anime countless of times
Since Google Colab became a paid subscription I decided to give one try with the free one, it took hours and then obviously the gen stopped.
With years passing by I lost motivation, so much so that I eventually deleted the zipfile of Cacao 6.1 LoRA and altough it might be available somewhere in bits, I would likely have to start all over again with prompts (but If I were to find it I promise I upload it as a zip file for anyone who knows how to create LoRAs correctly, as I was looking for tutorials even while making 4.0 so I was happy that so many people like it despite being under or overtrained.
For me to create 6.1 would recquire to either start over or have the patience and mindset to go through with this which honestly I don't
Last but not least, the model was suppose to be an SD2.5, and I don't know how relevant it is to use SD2.5 nowadays with XL, 3.5 being a thing.
Here's what wouldve been new in version v6.1
- Trained on 237 images instead of the previous 81
- Extreme focus on generating high-quality images
- Every outfit from the Nekopara anime is now suitable
- Every pose that Cacao had in the anime now trained in the LoRA
- Better prompting, focus on facial expression, variety enhancements
- 768x768 picture training was used instead of 512x512
- better focus on backgrounds
- Environments are now a key factor (meaning: objects, wind, indoors, outdoors working better)
- Fixed and minimized overtrained and undertrained images
- Captured the Nekopara art-style via training the Lora
- It is both trained on the AnyLora and AnythingElse V4 checkpoint models
additional information, generation guide will replace this textbox on release, this is just a tease on what to expect. Keep calm :3
Description
Trained on 81 images instead of 35
Bigger focus on generating high-quality images
Trained on more angles, positions
[Cat hat, no cat hat] is now working better
More outfits (kimono, pajamas, tank top, dress, overalls)
More facial expressions
I basically went through the whole anime and created a screenshot from almost every scene where Cacao appeared. This is currently the final version, if I ever going to create another version there will be with more images (supposedly 110) and I will train them in 768x768 ratio rather than 512 until then, have fun ^^
FAQ
Comments (15)
Update 6.1
[postponed to Tuesday]
So I finished collecting images for the final version. The total count is 237, I will crop them tomorrow then create and upload the actual LoRA Saturday. It will be suitable for almost every outfit that was in the anime but the best working one obviously be the overalls. Captioning this many images will be painful but I will try my best to create it.
(You can also expect it to be both suitable for AnythingV4.5 and AnyLoRA at the same time)
After that according to the Asahi Oka model I don't know when will I be making it but that probably don't need 200+ images, so I hopefully be able to start and finish V1.0 (which will be the first and only version) before I go on to my holidays. The reason the Cacao model needed this much is because Cacao has a tail, and also she is often seen with her iconic hat and stuff while Asahi really has her regular outfit and the dinosaur one with the Dream mask and stuff so I think I should be good with just 100. Stay tuned ^^
You have no idea how much I love this character compare to the other nekopara characters! Thank you!!
I kinda do since I chose her as my first ever LoRA material ._. Sadly she does not appear in the games.
@PepsiisgUWUd382 I know how you feel anyways good luck on your future Lora
I have to do over 80 more prompting and concept but after that the final version of the Cacao LoRA will be created and hopefully uploaded today or tomorrow at max. Stay tuned ._. (P.S this was a painful amount of captioning and training)
I wanted to finish but sadly it gotta wait after my holiday (which is next week Wednesday) I don't know when will I be ready I don't really want to give a date but somewhere around next week. I finished with the prompting I just have to run the tests and choose the best result and upload it but sadly that would take hours so I gotta postpone it til next week. Stay tuned ^^
cool Very good model
6.1 Update?!
Hey there! I wanted to give an update about this, did I shut down the 6.1 update or am I working on it? And the answer is yes, and yes. The problem is I just do not have the motivation to do this since I am creating LoRAs through the LoRA creating colab, and I have like 237 images for this update to train the model on and the way colab works is that if you aren't running anything it just shuts down causally, whether you pay for it or not while writing the tags, concepts for each image and I restarted countless of process until I think I found a solution in late July, which is creating txt and context files along with the zip, so I only have to upload the images and the tags will be there immediately without me have to rewrite, update, describe any images. This solution is working, I tested it out with simple 50 images, but my motivation cause of the failed attempts dissapeared when it comes to creating LoRAs
I never touched the project since early August, but now I said, okay let's finish this and started to continue tagging (which is basically copy-pasting from the previous text files into 2 different ones), I don't want to give an update on this, but I want to say September or early October will happen if I get the right motivation on creating this, since this is just a random and odd hobby for me, which I wanted to try. After the Cacao model is done I still want to create the Asahi model but then I might stop with LoRAs after that.
P.S: This Cacao LoRA could be a checkpoint too with the amount of poses, date, outfits it's trained on.
Like me, you're a person who wants to train a model to the extreme. Come on
I have used your model to generate several pictures and posted them to pixiv without making profits. I hope this will not affect your rights and interests. If so, please contact me and I will remove them
I don't mind it at all, so feel free to use it ^^
@PepsiisgUWUd382 Ha ha cool~
final update before release: I postpone the update with 2 days, because I trained the model, but it needs further improvement, and need to test it out, since the current one looks undertrained, so I wouldn't post that since it would feel like a downgrade compared to the previous version so stay tuned and thanks for your patience ^^
ALSO A SAD NOTE probably for some:
Cacao Version 6.1 will be my final LoRA model. The reason why is not the amount of work but colab. I could run SD locally too but sadly it would take ages since I have an AMD GPU and it is not recognized by Stable Diffusion so I turned to colab when I created my first two versions of the Cacao model like I said in one of the comments before. This was stressful mainly because the timeline suddenly disconnected during the training and now Colab started to kill 99% of Automatic1111 and anything that is Stable Diffusion related so I can't really test if the LoRA is good only for maybe 1 or max 3 images and it is sad for me to say because as much as I struggled, worked on LoRAs day and night, I loved creating them cause of the feedbacks I received and it was a new and kinda fun hobbie that I tested myself in. So until I found a method that AMD GPUs work with SD the LoRA creating journey of mine suspends after the Cacao ver6.1, it's even a miracle that I still haven't give up on this because it turned into a pain. so thank ya for understanding and much love <3 I try to release the 6.1 LoRA on Monday as soon as possible and I don't want to postpone it any longer.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.