v1-5-pruned-emaonly - CivArchive (CivitAI Archive)

v1-5-pruned-emaonly - v1-5-pruned-emaonly

Stable Diffusion is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input.

Description

FAQ

Comments (12)

YulexuanApr 22, 2024· 5 reactions

CivitAI

Interesting to see people still using the original V1.5 model to this day, even though the community has extensively fine-tuned it. I still use it for Lora training as my base model, literally every merge contains it in some way.

civitaidirectly068272May 12, 2024

What should people use if not this one? I'm new to training, looked up a few guides, but none of these tell you what to use as a model or model type or as a "source checkpoint???". I'm feeling pretty lost ...

YulexuanMay 14, 2024

@civitaidirectly068272 you can use another model like AbsoluteReality or RealisticVision, but compatibility with all checkpoints is not guaranteed, none of my Loras trained using mixes came out well. You'd like to use a model from which all others were trained on top of.

ScratchyKatAug 14, 2025

It is the model you get when you download A1111, like me., and I do think it's a great intro to AI image generation, even if not that great today. Personally, I use UltraRealism by Stable, Anything V3, and Dreamshaper8 LMC, primarily, but I'm pretty much always experimenting with new models

mycivitai512May 10, 2024

CivitAI

This is terrible. It cuts the heads off of people almost every gen. I say full body, it gens a person with no head, head out of frame. Why would anyone train a model on headless people? Use full body, face, side face, back, why headless? Never cut the head off the people please.

YulexuanJun 4, 2024· 7 reactions

This is the base model of all others that have emerged. These blurry outputs started the generative AI revolution we're now used to. I'm grateful it existed <3. It is now a piece of history and a model recommended just for training.

littlefluffyballDec 8, 2024

you must use portrait image size, and it will not cut heads.
It is probably the first confusing thing every SD user encounters.

littlefluffyballDec 8, 2024

@YulexuanThe question was how to fix head crops, not your philosophical essay.

YulexuanDec 9, 2024

@littlefluffyball valid, but almost no one is using it by now, the finetunes are better. This is only useful if you want to train loras in a way that will work with all finetunes based on this.

NourdalMay 17, 2024· 46 reactions

CivitAI

SD1.5... Yes, it is crooked, oblique, and completely ugly and uncomfortable. BUT... You have to RESPECT it!

Why, you ask? And I will answer.

Because it's a MOTHER! And, as with a real mother, you may not like her weird pie or spiky knitted socks, but you have to respect and say THANK YOU!

Just remember that all those cute SD 1.5 models and merges that you are currently using, generating fantastic landscapes, seductive girls or brutal warriors... they came out of the womb of this model. And there are THOUSANDS of them!

So, you don't have to use it, but... just show SOME RESPECT. Stop your usual generation, and try to make a new picture in it, at least one.

Believe me, she will be very pleased that you remembered how it all started. ;)

HardyBiggAug 7, 2025

Aw, and yet no one every sends her a Mother's day card. Ungrateful brats.

MelynsJun 18, 2024· 32 reactions

CivitAI

The worst model you can possibly download.

Checkpoint

SD 1.5

by faye_1225322

Download (Beta) View on CivitAI

base model

Details

Downloads

34,869

Platform

CivitAI

Platform Status

Available

Created

5/10/2023

Updated

5/26/2026

Deleted

Trigger Words:

cat

Files

v15PrunedEmaonly_v15PrunedEmaonly.safetensors

Size:

3.97 GB

SHA256:

6ce0161689b3853acaa03779ec93eafe75a02f4ced659bee03f50797806fa2fa