13 Model Comparisons for Stable Diffusion & Automatic 1111

Today I am comparing 13 different Stable Diffusion models for Automatic 1111. I am using the same prompts in each one so we can see the results side-by-side. Be sure to watch till the end, because there are some surprises.

[@aijohnart]

Positive Prompt: RAW photo, (masterpiece, best quality:1.4), realistic, full body, highly detailed skin, A serene woman with delicate features sits gracefully on the edge of a cozy bed, bathed in the warm, golden hour morning light that streams through a nearby window. Her soft, wavy hair cascades gently around her shoulders, framing her face. She wears a short, closed, luxurious silk robe adorned with an intricate Asian print, featuring blossoming flowers and delicate birds. The robe's vibrant colors and patterns contrast beautifully with the calm, neutral tones of the bedroom. The sunlight casts a soothing glow on her face, accentuating her expressive eyes, gentle smile, and flawless complexion. The peaceful scene evokes a sense of tranquility and relaxation, trending on ArtStation, 8k uhd, dslr, soft lighting, high quality, film grain, Fujifilm XT3 Negative Prompt: Double Body, blurry, logo, watermark, signature, cropped, out of frame, worst quality, low quality, jpeg artifacts, poorly lit, overexposed, underexposed, glitch, error, out of focus, (semi-realistic, cgi, 3d, render, sketch, cartoon, drawing, digital art, anime, manga:1.3), amateur, (poorly drawn hands, poorly drawn face:1.2), deformed iris, deformed pupils, morbid, duplicate, mutilated, extra fingers, mutated hands, poorly drawn eyes, mutation, deformed, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, incoherent Settings: 512x910 - CFG 11.5 - Sampling Method: DPM++ SDE Karras - Restore Faces turned on Models Used: Anime Abyss Orange Mix civitai.com/models/9942/abyssorangemix3-aom3 Perfect World civitai.com/models/8281/perfect-world ReV Animated civitai.com/models/7371/rev-animated Photorealistic Models: 2D fake disassemble 2D civitai.com/models/41294/2d-fake-disassemble-2d BRA(Beautiful Realistic Asians) V5 civitai.com/models/25494/brabeautiful-realistic-asians-v5 ChilloutMix civitai.com/models/6424/chilloutmix CyberRealistic civitai.com/models/15003/cyberrealistic Deliberate civitai.com/models/4823/deliberate Dreamscapes & Dragonfire (DNW 3.0) civitai.com/models/50294/dreamscapes-and-dragonfire-dnw-30 Food Photography civitai.com/models/45322/food-photography LOFI civitai.com/models/9052/lofi RealBiter civitai.com/models/16592/realbiter Realistic Vision V2.0 civitai.com/models/4201/realistic-vision-v20

[@user-jl2cm6ox6d]

Thank you for this comparison... what I noticed negative on some models is the badly formed limbs especially for BRA (V5) his left knee seems deformed, crushed for some other models are the hands which sometimes seem small compared to the body or even arms... In any case, thank you very much for giving us this opportunity.

[@midnightstories5695]

This is great stuff save us a lot of time testing Thanks!

[@AmeenAI4.0]

thank you for the effort amazing job

[@Seethesmall]

great video ! hope to saw more model comparisons!! thanks your tutorial!! it very useful! hope you will test the Lora model!!! or animate model!

[@luxo117]

thx this was useful , being a noob in AI I really appriciate this

[@davidjacobs8558]

some of these are checkpoints and others are loras I don't think loras work without checkpoints, so majority of the image quality is derived from whichever checkpoint being used.

[@aijohnart]

Hi David, I didn't use any loras with these models. I wanted the comparisons to be as fair as possible.

[@davidjacobs8558]

@@aijohnart but food photography and fake disassemble are lora lycoris

[@iREM0015]

the stable diffi not work now iam sad 😢😢😢😢😢

[@davidjacobs8558]

can you do this kind of video with "blonde European female" showing them standing up with full body from head to toe shown?

[@aijohnart]

Thank you for the suggestion. I know getting a full body image can be tricky. One tip I have is to describe their hair and their shoes. That way, stable diffusion will try to show you both, giving you a full body image.

[@davidjacobs8558]

@@aijohnart Dreamshaper v5 seems to be able to show full body better than most other models, in my experience.