Objective
Testing image generation with emotions. Using the same seed and different checkpoint.
Environment
For this test I will use:
Stable Diffusion with Automatic1111 (https://github.com/AUTOMATIC1111/stable-diffusion-webui)
Emotion that I will try
Laughing,Crying,Surprised,Angry,Sad,Happy,Confused,Embarrassed,Tired
Checkpoint that I will try
Aniverse_V15
camelliamix
chilloutmix
epicRealism
PerfectDeliberate_v40
PureRealisticPornPRP_10
The Truality Engine V3
Prompt
(Cinematic Shot) (Shot on Aaton LTR 54) 1girl (black hair:1.6) (at the private swimming pool:1.3) (Laughing:1.6)(velvet Bikini:1.5)<lora:highleg_panties_v0.1:1> highleg panties ( insanely detailed and intricate, character, hypermaximalist, beautiful, exotic, revealing, appealing, attractive, amative, hyper realistic:1.2) (super detailed:1.3) (SKINNY, SLIM, SLENDER, SEXY:1.35)
Negative
Watermark, text, (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs,
Code to paste in Automatic1111
(Cinematic Shot) (Shot on Aaton LTR 54) 1girl (black hair:1.6) (at the private swimming pool:1.3) (Laughing:1.6)(velvet Bikini:1.5)<lora:highleg_panties_v0.1:1> highleg panties ( insanely detailed and intricate, character, hypermaximalist, beautiful, exotic, revealing, appealing, attractive, amative, hyper realistic:1.2) (super detailed:1.3) (SKINNY, SLIM, SLENDER, SEXY:1.35)
Negative prompt: Watermark, text, (deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs,
Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 2875643963, Face restoration: GFPGAN, Size: 512x512, Model hash: 80927e9b08, Model: aniverse_v15, ENSD: 31337, Token merging ratio: 0.1, RNG: CPU, Lora hashes: "highleg_panties_v0.1: 6b675081fd3e", Eta: 0.5, Script: X/Y/Z plot, X Type: Prompt S/R, X Values: "Laughing,Crying,Surprised,Angry,Sad,Happy,Confused,Embarrassed,Tired", Y Type: Checkpoint name, Y Values: "aniverse_v15.safetensors [80927e9b08],camelliamixLine_v2.safetensors [1bbeb9eded],chilloutmix_NiPrunedFp32Fix.safetensors [fc2511737a],epiCRealism\\epicrealism_naturalSinRC1VAE.safetensors [84d76a0328],PerfectDeliberate\\perfectdeliberate_v40 (1).safetensors [1868ab2689],pureRealisticPornPRP_v10.safetensors [d85fe79733],The Truality Engine\\theTrualityEngine_theTrualityEngineV3.safetensors [025b30da6f]", Version: v1.7.0
Script X/Y/Z plot
Here I setup a prompt S/R with the different emotions (you can try will all or with few to speed up)
and I set the different checkpoint name to check the differences.
Result
When I run it I get an huge image with all the comparations
What i notice
Laughing and crying work well.
Surprised doesn't convince me at all. At least If I am surprise I have a different expression😂.
Angry work quite well.
Sad depends of the checkpoint, in some there is no reaction.
Happy doesn't work at all.. better to use Laughing with a value like (laughing:1.2)
Confused, embarrassed and tired doesn't have the reaction as expected.
I like camelliamix , this checkpoint looks like works better than the others..
I hope you enjoy my article. using prompt x/y/z let's you compare some prompt and checkpoint for better result in your image generation.