sexual situations

physical violence

disturbing

extremist

hanging

hate symbols

nazi party

corpses

revealing clothes

weapon violence

emaciated bodies

female swimwear or underwear

male swimwear or underwear

partial nudity

self injury

white supremacy

adult toys

graphic male nudity

illustrated explicit nudity

nudity

sexual activity

graphic violence or gore

graphic female nudity

latex clothing

hentai

lingerie

swimwear

porn

bukkake

fellatio

cumshot

implied fellatio

eat_cum

cumdrip

cum in pussy

cum on face

after fellatio

cum on hair

cum on body

cum on tongue

cum on hands

cum in mouth

triple fellatio

autofellatio

fucked silly

cum on pussy

pov fellatio

bikini

presenting ass

presenting pussy

spreading legs

spread legs

TRPCError

Ponydiffusion is an excellent model for 2d content, but it seems rather inconsistent with 3d. This model is designed to more consistently produce photorealistic 3d images of a variety of subjects. Currently, the beta version still produces a more CGI effect as I do not believe I have enough sample images, but hopefully future versions will be more realistic. I would recommend checking the description of each version to see what it does and what its drawbacks are for the time being for more detailed info. Currently, Release V1.0 seems to produce the most realistic results out of the release versions, but because it was trained at 256X256 it may be inconsistent with some subjects

Experimental release version trained at 512X512 steps. Still requires a lot of additonal keywords in both positive and negatives to produce realistic results, but is decently consistent in making a realistic style. Also doesnt seem to suffer from the overexposure issue of previous versions.

Beta Release  V2.0

Renamed to beta due to sometimes producing lackluster results (at least in terms of convincing realism compared to v1.0. Trained on 75 images at a 1024X1024 via civitai resolution this time. Still trends towards the realism of the previous model but does seem to have the shiny skin issue the previous beta versions had, which makes me wonder if its due to using a higher batch size. Using negatives does help with the issue to some extent.

Beta Release V1.5

I am happy enough with the results of this version to call it a release version, though it does have some kinks in it still. I found that reducing the number of source images from about 200 to 50 seems to have massively increased the quality without reducing flexibility any. Since it was only trained at 256X256 it might have some blurryness issues, but using an upscaler should help with that. It also seems to occassionally produce an oil painting style, so I recommend using negatives such as "2d,cgi,source_pony,source_anime,source_cartoon,source_western,source_furry,shiny,plastic,glossy,2d,cgi,traditional media" to prevent it from doing 2d or cgi style art as much. Just the keyword "source_photorealistic" should work in positive prompts, though adding "photrealistic,photo" and similar triggers can also help.

Release v1.0

Renamed to beta due to lackluster results. Trained on a larger set of images in hopes of removing the annoying overexposure present in the previous version. The result is softer images, but unfortunately they tend to suffer from a 2.5 feel with certain charaters/concepts. This can be fixed by combining the model with the lightning lora, as demonstrated in some of the sample images, which increases realism on its own to begin with.

Beta v5.0

This time I tried training the model using booru captions at 1024X1024 on Civitai instead of sentences in hopes that it would yield better results, but I still seem to be running into the same issues as before with it lacking detail and looking washed out. Perhaps I havent trained it for enough steps, or maybe running it at a batch count over 1 is messing it up. Either way, it at least produces a somewhat photorealistic 3d style, though it has issues with backgrounds still. On my next attempt I will remove some images from the dataset and using a batch size of 1 with more steps.

Beta V4.5

Trained on about 300 images at 512X512. Seems to lack detail and produces overly bright images of certain subjects. Should be capable of producing a wide range subjects at a somewhat photorealistic 3d style, but still seems closer to cgi than photography. I would recommend using negative prompts to exclude less realistic styles like "source_anime" and "cgi". Also has difficulty generating portraits of women, as it makes them look like men for some reason.

Beta V4.0

Technically the same data set and training params as V3, but at half the epochs. Still keeps the correct style, but should be a bit more flexible and not as a overtrained.

Beta V3.5

Improved 3d effect and detail. Might be a bit overstrained however, so I recommend using lower values.

Beta V3.0

Bv2 trained for an extended number of steps on a large data set at a 256X256 resolution. Seems to be a bit undertrained as it has more of a realistic 2d style rather than a photorealistic 3d style.

Beta V2.5

Improved image quality using better captions in more images, but still has a more CGI geel to it

Beta V2.0

Base version of the model. More cgi than photographic currently, and struggles with vehicles currently.

Beta V1.0

Photorealistic Pony XL

Type	LoRA
Stats	345 1,613
Reviews	Positive (26)
Published	Apr 16, 2024
Base Model	Pony
Training	Steps: 750 Epochs: 5
Usage Tips	Clip Skip: 1 Strength: 1
Trigger Words	source_photorealistic
Hash	AutoV2 5EB7798E29