NaughtyPuff
patreon
π€ Naughty Academy: #1 β Choosing Your Model & Setting the Perfect Configs
🕑 Added 2025-05-13 22:06:21 +0000 UTCWelcome to the first step of the NaughtyPuff Academy!
In this post, I will guide you through the beginning of the process: choosing your base model and settings.
I'm focusing 100% on my personal workflow, explaining exactly what each step and setting means, so you fully understand what you're doing β not just copying settings blindly.
β οΈ Important Note:
Due to Patreon's policies, I cannot mention specific tools or platforms.
You can generate images locally (if you have a good PC) or online through various services.
There are many tutorials on YouTube if you want to learn about local setups.
π― Step 1: Choosing Your Base Model
The first step is to choose your main model.
This model contains all the fundamental knowledge needed to generate your images.
Here are two excellent models I recommend:
β Option 1: Hassaku XL (Illustrious) v1.3 - Style A
Delivers a clean and soft visual style.
More minimalistic and elegant.
β Option 2: WAI-NSFW-illustrious-SDXL (currently v14)
Currently my main model.
More anime-focused by default.
Easily adaptable using LoRAs (which we'll discuss later).
π§ Why Choose Illustrious-Based Models?
There are many different models available, but Illustrious-based models stand out because:
They are more consistent in results.
They follow prompts better.
They offer higher quality and more flexibility.
Tip:
I recommend avoiding Pony-based models, as they are a bit outdated and less responsive to complex prompts.
βοΈ Step 2: Basic Generation Settings
Each base model usually has its own recommended parameters.
Hereβs what I personally use, along with a simple explanation of each setting:
π§ͺ Sampler: Euler A
Controls how the model interprets your prompt and builds the image.
Euler A is great for fast, smooth, and detailed results.
Tip:
You can experiment with other samplers later, but Euler A is my favorite for consistency.
ποΈ CFG Scale: 6
"Classifier-Free Guidance Scale."
Controls how strictly the model follows your prompt.
Higher values = more accuracy but less freedom.
Lower values = more creativity, less strictness.
My preference:
I usually set CFG Scale around 5β6.
Sometimes even 4.5 for a softer, more artistic touch.
β±οΈ Steps: 25
Defines how many refinement steps the model takes when building the image.
More steps = higher detail, but also longer generation times.
Ideal range:
For my workflow, 20β30 steps is more than enough, especially when using Euler A.
β Quick Recap:
Choose a good base model (preferably Illustrious-based).
Use basic settings to control the generation (Sampler, CFG, Steps).
Comments
CaptainDeadDaddy
This is awesome!
gerbil56
Will we ever see more of these? It would be awesome!
Ricky
This post is filled with very useful guidance. I look forward to more. Thank you for sharing your knowledge!!
T3rrymcg1nn1s
Really awesome of you to do this!!
