You can use my model directly at : Tensort.Art | Mage.Space
_____
Units can contact me if they want to release for commercial or exclusive use ( Facebook | Zalo | Mail )
_____
If you find it useful, please vote 5 stars and post an image to help me, it's really important, I thank you all very much.
_____
In addition, I have some LoRa trained for SDVN8, you can refer to it on my TensordArt page.
_____
📖 Checkpoint Info
Info : I'm happy to have completed my best checkpoint collection to date. SDVN10-Anime is a collection of great checkpoints in anime style. I have trained further and adjusted using many different methods to create upgraded versions. You can see more at the official website to see the greatness of this checkpoint set. Currently I provide the base 1.0 version for free at CivitAI, you can try all other versions at Tensord.Art or download all versions for $5. I hope that everyone can support me so that I have more motivation to develop new versions of this checkpoint set
____
You can refer to some of the indicators below to achieve the best image quality :
CFG : 7-10
Steps : 20-32
Size : 512x768 px ( or 640x960px )
Sampling : Euler a
📝 Quality Prompt :
Base quality prompt :
(masterpiece,best quality:1.3)Key :
ghibli, arrietty, castle in the sky, from up on poppy hill, ghibli, grave of the fireflies, howl_s moving castle, kikis delivery service, my neighbor totoro, on your mark, only yesterday, pom poko, ponyo, porco rosso, princess mononoke, spirited away, tales from earthsea, the ocean waves, the wind rises📝 Negative Prompt :
watermark,text, logo,contact, error, blurry, cropped, username, artist name, (worst quality, low quality,nsfw,nipple, pussy:1.3)_____
Also you can test those prompt tags quickly with my model here : Tensor.art
I have written a colab site that integrates all tools for you to use stablediffusion without configuring your computer, you can refer to : Colab SDVN
I have provided an album with more than 2000 original photos with full prompt, you can refer to understand how to use SDXL : SDVN-SDXLPrompt Kit
In addition, some of my models are available on the Mage.space platform, you can refer to: SDVN Mage
_____
If you find it useful, please vote 5 stars and post an image to help me, it's really important, I thank you all very much
_____
👨🏻💻 Creator : stablediffusion.vn
📢 Discord : Discord SDVN
📢 Facebook : Group SDVN
☕️ Donate me : Bank | Buymeacoffee
Description
This is a version trained in Ghibli style, my favorite animation studio.I train completely at 1024px, so it works well even at 640x960.
FAQ
Comments (13)
A question out of curiosity, if focusing on specific styles, wouldn't it be better to make LoRAs instead of full checkpoints?
With tensor rt you can't use loras so it's good to see checkpoints as well
as someone that loves these checkpoints that I can merge with other models, no it wouldn't be better.
You can just merge LoRAs into checkpoints very easily and it's much more flexible to use. In fact I extracted a LoRA from this model and it works very well with multiple models. It's easier to store many LoRAs of few MBs and use what is needed than many checkpoints of ~2GBs. LoRAs are much easier and faster to train, so probably there could be more styles available more easily.
@victorc25744 sure, if you use A1111, I can't unfortunately. I can train loras, and I can merge models, but I cant merge loras with models
@Fenn I don't use A1111
I train this style with lots of images, in this case, checkpoint will achieve better results
@StableDiffusionVN I see. And after training did you merge back layers from the Artemis model? There are some of the cross attention layers that are identical to the ones from Artemis v4, I was curious about that
@victorc25744 This is the original model trained from Artemis, I did not influence any factors such as merging more models or lora
@StableDiffusionVN what did you use to train? It's not that there's anything wrong with it, but I'm trying to understand why those layers are identical, just to see if training can be optimized somehow :D
amazing
From the examples it seems this is just Studio Ghibli only?
I love Ghibli style! Unfortunately, despite the fact that this model is very cute, it can be unstable and buggy in some topics.
But I still really like it! :) Cool style! Thanks!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.