Hello everyone!
All models are exclusive to Civitai ! Anyone who publishes my models without my consent will be reported!
Trigger Word : Anime style
I highly recommend loading my lora models in fp16 if possible
Since Flux's anime style is a bit weak, I decided to tackle it for my first training of Flux LoRa model.
v3_Final : This version uses the same dataset as v3.1 but this time at 100 epoch against 16 before.
The poses are more dynamic and the characters more expressive
V3.1 : A new dataset of 330 images. Better selection so that characters look older if requested.
The overall quality is also much better
V2_Final : same as v2_15k, but the style is much more pronounced
v2_15k : This version has a dataset of 338 images against 112 for version 1.
This is not the final version of this v2, but I find this 15k steps version very interesting and rather stable.
This version was trained with the optimizer "prodigy" in rank 32
The style is there, but the final version will be very different!
V1 : I used for this 112 high quality images generated with MidJourney v6 in the --personalize style.
Be indulgent, this is a first try. If you zoom in a bit on the images the lines can appear a bit pixelated. But the style is pretty cool!
For a pronounced anime style, Sweet spot between 0.90 and 1.30
I recommend you to play with the strength to vary the pleasures for example at 0.5 or 0.6 by asking for a realistic image.
The model was created with Ostris script which you can find here :
https://github.com/ostris/ai-toolkit?tab=readme-ov-file#flux1-training
Have fun! 😊
Description
FAQ
Comments (36)
I love seeing this! Great job!
Thank you very much. It's still experimental, but I'm quite satisfied 😊
@Stan_Katayama Proving those who said this could not be trained wrong!
@crawlingrat There are some who always have an opinion on everything day 1 without knowing anything at all 😅
@crawlingrat And thanks for the buzz 😊
@Stan_Katayama Np, gonna be sure to buzz as many Flux Lora creators as I can as soon I have enough. Ya'll are doing the lords work!
@crawlingrat That makes me very happy! But keep your buzz! I have some in reserve! What would make me happy, however, is if you published some images under my model. I love seeing what others do with it :)
insane work bro !!
Thanks a lot ! 🙏
What prompt do you use for the different styles of anime, artist names?
Fortunately the workflow isn't stripped from the image's metadata that the author posted.
In the description I say that I used images from MidJourney. Which implies that there is no style other than MidJourney. If you want anime style just write the keyword "anime style" :)
@veralapsa That's what I'm going to do next time! I'm going to publish a model, keep the little secrets of how it works to myself and upscale the images with third-party software so as to leave no trace of the metadata. 😂
Once again, the problem is that Civitai has trouble parsing ComfyUI workflow. So to see the prompt, you need to click on the download button above the image, then drop the PNG into ComfyUI to see it.
Once again, the problem is that Civitai has trouble parsing ComfyUI workflow 😅.
So to see the prompts for the sample gallery, you need to click on the image, then click on the download button above the image, and drop the PNG into ComfyUI to see it.
Am I dreaming or are you giving the technique to steal all my secrets? 🤔😂🤣
@Stan_Katayama I am all for sharing secrets 😂🙏😎
@NowhereManGo 😂🤣
bro, the lora is not changing any style for me i dont know whats wrong.
Have you updated comfyui? With the arrival of Flux you have to update very regularly.
Yeah, what he said. If you don't update comfy it literally has no effect.
@civitai426 yeah, it works now thanks
I am getting this error when using this LoRA on Forge: [LORA] LoRA version mismatch for KModel: F:\Forge\stable-diffusion-webui-forge\models\Lora\MJanime_flux_lora_v1_000004000.safetensors
I tested on Comfy and Forge and both either throw an error or just don't load. I did make sure to update Comfy.
This problem was fixed very quickly on comfyui. It will most likely be fixed on forge too. Try to update it regularly
@Stan_Katayama Thanks for your reply! I will keep an eye out for Forge updates.
@cubby_ai I hope it happens quickly.
Let's Gen It!
Flux dev fp8_e4m
This is my favorite LoRA to date. I'm not sure if that matters because stuff changes so fast.
Flexibility: Works with other loras easily and has a high range of control in accordance to guidance and strength.
Adherence: It's a little more difficult with getting prompts down. However, it's always suprising to see what it comes up with. The results are quite provocative to the imagination.
Consistency and Variation: This lora appears to be quite consistent when sticking with the same seeds or even the same prompt. But variations are not to extreme from one result to the next.
thanks for this very complete feedback! 🙏😊
Make one for animagine xl or pony diffusion
I will try. But the results would be far from this quality.
@Stan_Katayama yes please try
Simply amazing and easy to use. 10/10.
Many thanks ! I am filled with joy 🙏
Amazing, does that work with nf4 ?
I don't know. I think so. You just have to put the strength a little bit higher between 1.0 and 1.3 I think.
AFAIK, Flux-dev LoRAs should work with NF4 if you have the latest Forge.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.