========================================================

Illustrious NoobAI V1.0

After making a few versions of NoobAI checkpoint, I finally find one that is an upgrade from the Illustrious version V2.0.

In short:

• Both Euler and Euler A can be used as sampler (others can be used too, but you will have to test and find out what setting work best for you)

• Use Euler and low CFG scale (1.5-3) if you want realistic looking image

• Use Euler/Euler A and high CFG scale (3-7) if you want 2.5/anime looking image.

(Fig.1 Difference on realism depending on CFG scale, running Euler as sampler)

If you have the time and interest to know a bit further:

Euler and low CFG scale (1.0) will give you very realistic image, like a photo, and hence you will have less flexible background, pose and lower saturation; CFG scale 2.0 seems to be a good balance between realism and artistic for most of the cases.

You can also pair with my dynamic lighting lora to enhance the lighting/saturation on realistic images, I find the results very satisfying.

Artistic tools [IL] - Dynamic lighting v1.0 | Illustrious LoRA | Civitai

(Fig.2) Examples of low CFG (2.0) and Euler with dynamic lighting lora

On the other hand, Euler/Euler A and normal CFG scale like 4-7 gives you results that is close to Illustrious version V2.0. But again, other loras/character used will affect the degree of realism, so you will have to play around and see what fits you best.

(Fig.3 Examples of normal CFG (5.0) and Euler A)

If you find it too troublesome, you can still pair it with my realistic filter lora, so you can control realism with the weight of this lora single handedly.

Realistic filter [IL] - IL v1.0 | Illustrious LoRA | Civitai

Recommendation:

Positive prompt: masterpiece, best quality, realistic

Negative prompt: worst quality, old, early, low quality, lowres, bad hands, bad fingers, mutated hands, greyscale, text, copyright name, logo,

Sampling method: Euler, Euler a

CFG: 1.5-8

Size: 832x1216

(Feel free to try out other settings)

Merged Models

NoobAI-XL (NAI-XL) - V-Pred-1.0-Version | NoobAI Checkpoint | Civitai

initiumFinetuneNoobai_V10 (unavailable on Civitai)

RealCosplay (or: noobReal ) - v2.1 (noob) | Illustrious Checkpoint | Civitai

Uncanny valley - UncannyValleyilxl1.0+noob | Illustrious Checkpoint | Civitai

Data from yomama 2.5D [Illustrious] [Pony] - Illustrious v2.0 | Illustrious Checkpoint | Civitai

License

Fair-AI-public-1.0-sd - This license disclaim "Same license restriction" so I do not add any restriction except same license restriction.

This model permits users to:

✔Use the model without crediting the creator

✔Sell images they generate

✔Run on services that generate images for money

✔Share merges using this model

✔Sell this model or merges using this model

❌Have different permissions when sharing merges

========================================================

Illustrious version V2.0

The realism of characters varies vastly, depending on the data of that character; e.g. manga/anime characters will be much less realistic than a 3D game character. I kept the weight of the realistic checkpoint low so the already realistic characters wont become still image, but that also means the manga/anime characters will be still slightly unrealistic, in that case you can pair it with my realistic filter lora which works very well with this v2 ckpt.

Realistic filter [IL] - IL v1.0 | Illustrious LoRA | Civitai

Because of the high weight ratio of the v1 ckpt, this v2 ckpt will mostly perform similarly.

Recommendation:

Positive prompt: masterpiece, best quality, realistic

Negative prompt: worst quality,bad quality,bad hands,very displeasing,extra digit,fewer digits,jpeg artifacts,signature,username,reference,mutated,lineup,manga,comic,disembodied,cropped,amputee,bad foreshortening,what,guro,logo,bad anatomy,bad perspective,bad proportions,artistic error,anatomical nonsense,amateur,out of frame,multiple views, text ( copyright name:1),

Sampling method: Euler a, DPM++ 2M karras.

CFG: 6-8

Size: 832x1216

(Feel free to try out other settings, if the results are too 2Dish, you can tune down the weight of the anime lora)

Merged Models

Illustrious-XL - v0.1 | Illustrious Checkpoint | Civitai

Evermore Illustrious XL - v1.2 Fp16 | Illustrious Checkpoint | Civitai

yomama 2.5D [Illustrious] [Pony] - Illustrious v1.0 | Illustrious Checkpoint | Civitai

License

Fair-AI-public-1.0-sd - This license disclaim "Same license restriction" so I do not add any restriction except same license restriction.

This model permits users to:

✔Use the model without crediting the creator

✔Sell images they generate

✔Run on services that generate images for money

✔Share merges using this model

✔Sell this model or merges using this model

❌Have different permissions when sharing merges

========================================================

Illustrious version V1.0

Recommendation:

Positive prompt: masterpiece, best quality, realistic, 2.5D, photorealistic,

Negative prompt: worst quality,bad quality,bad hands,very displeasing,extra digit,fewer digits,jpeg artifacts,signature,username,reference,mutated,lineup,manga,comic,disembodied,cropped,amputee,bad foreshortening,what,guro,logo,bad anatomy,bad perspective,bad proportions,artistic error,anatomical nonsense,amateur,out of frame,multiple views, text ( copyright name:1),

Sampling method: Euler a, DPM++ 2M karras.

CFG: 6-8

Size: 832x1216

(Feel free to try out other settings, if the results are too 2Dish, you can tune down the weight of the anime lora)

*This version is quite good with non-human or weapon generation. Please let me know if there is anything wrong, as always, have fun and go break some eggs.

========================================================

Pony version

Dear all, it has been a while since I last updated, I had been experimenting with different things, and this is the first time I upload a checkpoint. Honestly, I have no idea what I am doing, there could be some mistakes, but I like the art style and I don't see it on civit, so I would like to share.

Recommendation:

Positive prompt: score_9, score_8_up, score_7_up, realistic/anime,

Negative prompt: score_6,score_5,score_4,worst quality, bad hands, bad fingers,

Sampling method: Euler a, DPM++ 2M karras.

CFG: 6-8

Size: 868x1312

(Feel free to try out other settings, if the results are too 2Dish, you can tune down the weight of the anime lora)

*This version is not good at non-human or weapon generation. Please let me know if there is anything wrong, as always, have fun and go break some eggs.