Haruno Sakura (Naruto - all series) LoRA
Making models can be expensive. Do you like what I do? Consider supporting me on Patreon 🅿️ or feel free to buy me a coffee ☕
There was already a LoRA for this, but people still requested it. I've asked Satyam (the author of the first Sakura LoRA) if that's fine for me to post this. If he somehow decided that it's not cool I'll simply remove it.
Back to the LoRA, this can do all versions of Sakura, from her child version with long hair, to her Boruto adult version. It can also control her forehead mark by simply adding forehead mark to the prompt (yep, the one in the example is not inpainted). To control her version you can use descriptive tags + the tags of the series she's from (see the examples). I suggest around 0.6 weight and triggers with haruno sakura. You can also have her wear any outfit you can think of simply changing the prompts. Examples made with AnyPastel and AbyssOrangeMix2. One is using my Sam Yang LoRA.
How to use LoRA's in auto1111:
Update webui (use
git pulllike here or redownload it)Copy the file to
stable-diffusion-webui/models/loraSelect your LoRA like in this video
Make sure to change the weight (by default it's
:1which is usually too high)
Description
FAQ
Comments (20)
Lykon onn thing I don't get is that you have all the defining characteristics in the prompt as well as the prompt tag. Do you have to include all those aspects with the tag as well?
do you mean during training?
so in your examples you use a lot of prompts as well as referencing your LoRA. its difficult for me to determine if the LoRA is being applied as when I use all of your example prompts the other descriptors are doing most of the work. how can I check if the LoRA is actually being applied correctly?
@muuuuuu Do one with the lora and then use the same seed to do other without it... I think, like I'm no expert but that's what I do
@muuuuuu people often get confused by this because they conpare with embeddings. Embeddings (ti) only store 1 concept, meaning if you train a "Super Mario" ti, you either get super mario as a whole or not. With lora that's a bit different. If I train with only 1 tag, so tripping away everything from the db, I can achieve with lora the same result, but that's not useful. I'd better do an embedding at that point, right? With lora I can tag every single part of super mario's outfit, allowing me to have super mario with hat, super mario without hat, super mario with the flower colors, super mario with the tanuki outfit, etc. Good examples of this are my Tifa lora and my Pyra/Mythra lora.
@luffon exactly.
@Lykon so could it work to combine all your loras into a single "Lykon AIO" lora
@friendfree6971 unless they have conflicting concepts. Eg: "pink hair" can mean one thing in a Sakura lora but something different entirely in the Melina one. Anyway I don't think that merging loras is useful as you can use any number of lora in the same generation.
this is very good, I hope you can also teach Ai to recognize the face of staz Charlie blood from the anime blood lad, he is my favorite anime character.
Are the checkpoint in the first six picturesusing AnyPastel?
It seems different from AnyPastel's painting style
Except the third one, but they're also using my negative embeddings. They tend to change the style a lot, making it much more controllable.
My review won't show up. Did you disable them?
Help, I got this error.
LoRA weight: 1, model: harunoSakuraNaruto_lykonV1(7449a011)
Error running process: /content/stable-diffusion-webui/extensions/sd-webui-additional-networks/scripts/additional_networks.py
Traceback (most recent call last):
File "/content/stable-diffusion-webui/modules/scripts.py", line 338, in process
script.process(p, *script_args)
File "/content/stable-diffusion-webui/extensions/sd-webui-additional-networks/scripts/additional_networks.py", line 177, in process
network, info = lora_compvis.create_network_and_apply_compvis(du_state_dict, weight, text_encoder, unet)
File "/content/stable-diffusion-webui/extensions/sd-webui-additional-networks/scripts/lora_compvis.py", line 73, in create_network_and_apply_compvis
network_dim = min([s for s in size if s > 1])
ValueError: min() arg is an empty sequence
Odd. Try to redownload maybe?
Hey Lykon, was just wondering, would you create more LORAs of Naruto characters, Itachi, Rock Lee, etc
if they're not made already, I might do it, but my list of lora to do is long
If someone would make a checkpoint based on Sunahara Wataru's art so I can use this I'd be so happy
Hi, can you do karin uzumaki too please.- your loras are so great <3
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.