Amaranth is primarily designed for Role Playing Game (RPG) character creations, but it can also do a lot of things like the general stuffs. So basically, it is also great for general use in the realistic or photorealistic image generations. I personally use this model for my works, which is well suited on one of my styles. Regarding text prompts, just be versatile and let your imagination flow. And, just like most models it also has few issues with the hands, which can be easily mitigated with this model's inpainting (download separately) or some embeddings. (I still have other models, for some specific arts, yet to be released. Your support will go a long way, and a cup of coffee will keep me awake for many nights. Thank you in advance!)

This version of Amaranth does not include a VAE because the goal is for art versatility, wherein, you can use any VAE you like to produce unique aesthetics. Below are the recommended VAEs that you can use with this model to tweak a unique output.

Popular Variable Auto Encoder (VAE)

kl-f8-anime2 VAE (for anime)

anything VAE aka orangemix VAE (for anime)

blessed VAE (for anime)

vae-ft-mse-840000-ema-pruned (for realism)

Sample images above were created with simple text prompts around default settings. More detailed prompts will produce much better results. As for text prompts, just let your imagination flow. One thing to take note, adding keywords like realistic, realism, photorealism, photorealistic, hyperrealism, hyper realistic etc. will make the image more real, and not including any word with real on them will make it look anime-ish.

Settings:

Sampling Method: Euler (realistic) or Euler A (anime)

Sampling Steps: 20 to 30

Resolution: 512x512, 512x768, 768x512

CFG Scale: 4 to 7

Popular Embeddings Used:

Easy Negative

Deep Negative

Bad artist Negative embedding

veryBadImageNegative

bad_prompt Negative Embedding

negative_hand Negative Embedding

EnvyBetterHands LoCon

Detail Tweaker LoRA

SXZ Vivid Darkness

Theovercomer8's Contrast Fix

Limitations:

Do use this model within the boundaries of legalities; do not use it to harm people and entities, or any illegal matters. Using this model means that you alone are responsible for any forms of consequences good or bad, and the author has no liabilities. We all know that AI models can produce both safe and unsafe images for work, so be very responsible. For all other matters do read the CreativeML-OpenRail-M license Other than that, enjoy using this model and I hope this will help you in your endeavours as an artist.

Support:

First of all, thank you for patronizing this model; it means my efforts didn't go to waste. My goal is to produce more free tools not just for myself and my work, but for others too. Your generosity will go more than a mile. Also, as an artist, writer, programmer, and overall freelancer, I accept commission job. Thank you so much for your support!