Back to Discussions
Chat

Is anyone still using Neta Lumina?

SSNoW
Translated to English·

I’m considering taking this model down and replacing it with another, more popular anime model.

I’m posting this because I want to know if there is anyone actually using this model extensively, rather than just trying it out. This model has clear strengths and weaknesses. Among the relatively new anime models capable of understanding natural language, this is likely the only one with a loose license.

However, it lacks sufficient training and isn't very popular, so there are few related tools and almost no optimization tools for it. As a result, it runs slowly, and achieving the same speed as Animagine requires more resources (= cost).

If I were to replace it with a different model, do you have any recommendations? I would appreciate your feedback, such as on Illustrious or Pony.

1

58 replies

Pinned
S
SNoWOP·

Regarding the model replacement, based on the feedback we recently received, we are currently considering the following plan. Instead of simply deleting "Neta Lumina," we are looking into switching to "Neta Yume," which is based on Neta Lumina and has undergone additional training with another 10 million images. We believe this may improve the issues with output stability that some of you have pointed out (at least based on discussions I’ve observed elsewhere).

We plan to keep both models available for about a week, and if there are no particular issues, we will proceed with the switch to Neta Yume.

👍3
I
ines·

You are right

宋
宋春阳·

Have

わ
わた個せ·

😆

た
たぬー·

I don't really use it as my main tool lately, just occasionally for comparing different expressions. I feel like it reacts to text in a straightforward, consistent way (with less fluctuation), but since V3 feels the same, I've been using that one more for testing.

❤️3
L
Leni·

Love

好
好家伙·
@Leni

Me too

L
Leo Wendt·
@好家伙

❤️

O
Otakosan·

I haven't been using it at all lately either.
I only make works with anime-style art, but since the accuracy is low and I don't like the default style, I've just been using Animagine.

감
감세안·

That's a good piece of writing.

V
ViaN_ ANIME·

Language is like a bridge.

こ
こばと🕊️_ダヴ(DOVE)·

It would be great if I could use Midjourney V7, but that's probably impossible, so perhaps Illustrious or Pony? 🤔 Does that mean I'd end up using LoRA? 🤔

ま
まれん2·

I used that a lot when I first started, but lately I've just been relying on ChatGPT.

さ
さな_全人類ケモナー化計画·

It would be pretty cool if we could use something like NoobAI.

👍1
V
ViaN_ ANIME·

Khonimoka sriko

や
やまびこ | yamabico·

I hardly ever use Neta Lumina.
It works sometimes for an art-style aesthetic, but for illustration-style, anim4 is definitely more consistent.
So, I would actually welcome a switch!
It's a bit embarrassing to admit, but I'm not that knowledgeable about models, so I can't really recommend something specific like "this one is good!" However, since I mostly generate illustration-style images, I'd be happy if I could generate illustrations with a different vibe. 🐔

👍3
S
SNoWOP·

We are planning to add two new anime-style models soon. They are Animij and WAI, both of which are part of the Illustrious series. We are preparing them as candidates to replace existing models.

👍5😄2
C
Christ amour·

Think carefully before making your decisions, in any case.

さ
さな_全人類ケモナー化計画·

It seems WAI released its latest version, v17, a few days ago, so I wonder if that's the one?

S
SNoWOP·
@さな_全人類ケモナー化計画

Yes, that's correct, v17! The other one is also a new version that was just released recently.

👍1
フ
フラン好きの時雨霊華·

I use this most often, or rather, this is the only one I use... I can only create images of the original characters I'm currently working on with this... Is it too much to ask if I could set the art style to a certain extent using prompts?

👍3❤️3
S
SNoWOP·

I'm glad to hear a positive opinion, as those are rare! I started this thread specifically because I wanted to hear everyone's thoughts, so thank you for commenting!
As far as I understand, Neta is quite good at art styles it has already been trained on, but it might be difficult to generate characters or art styles that it hasn't been trained on using just prompts.
Please feel free to refer to the two documents below:

  • Official Prompt Guide
  • NetaYume Art Style Keyword Collection (This is a model that further fine-tunes Neta, but much of it can be adapted)

You might find keywords that are close to the art style you are looking for.

👍1
V
ViaN_ ANIME·

Hello friend

N
NS0813·

I'm getting wildly inconsistent results; sometimes I can create perfect 2D anime images of dark-styled beauties, and other times they just don't work at all.
image

ぎ
ぎにゃー·

I don't use it very much. That's because when I try to get it to draw characters, the shapes get a bit distorted or, for some reason, two of them appear. When I do use it, I use it for things like landscape paintings.

❤️2
V
ViaN_ ANIME·

Good job, friend

い
いいじゃ·

I use it often. I still use it because there are compositions I can generate with it.
Rather, this model often doesn't perform well with ordinary scenes, but for some reason, when I give it sensitive prompts, it suddenly gets serious and often produces amazing illustrations.
That's why I don't want to part ways with this model.

❤️2
紫
紫苏·

Yes, there is.

·

I use it often!

V
ViaN_ ANIME·

We are going to the fields

G
Ggg790·

Me too, I prefer others because they beat it in the competition, and I like their style better.

こ
こばと🕊️_ダヴ(DOVE)·

I think you should just put it in the others section. Wouldn't it be fine to put it together with image v2 or SDXL? 🤔🕊️

👍3
ね
ねむ·

I don't think a replacement is necessary.
I feel that the images generated by Neta Lumina can be hit or miss, but through using it, I've found that images generated with prompts that are neither too short nor too long turn out well.
I've pasted an example of a success below.
image

👍5
V
ViaN_ ANIME·

Eat with your hands

❤️1
P
Pizzaiolo·

good

R
Ramon·

Yes, say

C
cosmos·

I use it often depending on the creations, the new apps are great too, congratulations to you. If possible, keep everything. Thank you very much!

❤️2
·

5555

C
cen·

I don't use it much, image is still more useful.

V
ViaN_ ANIME·

Khonaile nukany

一
一个菜鸟UP·

Oh

あ
あずきバーは飲み物·

I don't really use it much anymore either.

こ
こばと🕊️_ダヴ(DOVE)·

Sorry 🙇 I couldn't help but laugh at your log name 🤣🕊️🌷

D
Diệu Ngọc·

ok

최
최민·

GG

T
Thanh Huu·

😀😀😀😀

T
the·

👍🏻👍🏻👍🏻

图
图咪生成·

666

C
cosmos·

with a lot of patience and research we arrive at superb creations always bravo to you and a big thank you:

J
jun an·

Not very familiar with it

K
ko3jou·

I used to use it, but I stopped because the accuracy was too poor.

こ
こばと🕊️_ダヴ(DOVE)·

Neta Lumina is no longer used on this site. It has been replaced with NetaYume. Is it possible to delete or rewrite the old comments? It's confusing 😞🌀🕊️

S
SNoWOP·

It's fine to leave that as it is. I was thinking of intentionally keeping it as a sort of record of the opinions on this model at the time.

👍1😕1

Leave a reply

Join the community to share your thoughts and ideas.

Sign in to join the discussion