Rebranding my AI model due to a naming conflict. The new name "AutHuman", I hope you like it. (I couldn't change the filename, but you can rename it after downloading.)
Recommend using ADetailer.
Sampling method: DPM++ SDE, DPM++ 3M SDE Exponential, Euler a (for R+)
My preferred VAE is: https://civarchive.com/models/152040?modelVersionId=360299
V2.5 is a beta version aimed at better matching the prompt descriptions in the generated images. However, I currently don't have much time to test its accuracy. I would greatly appreciate it if everyone could download and test it, and provide me with some feedback.
The new version V3.0 aims to primarily enhance the background and lighting effects, with a focus on improving the depth and realism of shadows and highlights.
If you're interested in exploring more of my experimental models, please feel free to visit my Tensort.Art profile page.
Description
FAQ
Comments (18)
There's already a model called Real Pony, why would you call it this lmao
My apologies, I failed to notice that issue. Thank you for bringing it to my attention; I will endeavor to correct that name appropriately.
@chatgpt2023126 If it helps, your model looks much more REAL than Real Pony.
@Drakeni Thank you for your input. I've always hoped it could appear more realistic, but that may come at the cost of sacrificing some of the original pony model's characteristics. I hope to strike a balance between realism and retaining those core pony features in future iterations.
How does it fare with males? I'll likely give it a go myself, just wanted to see if you have any input as well.
In terms of interacting with males, it retains the key characteristics of the pony model. However, I must admit that it cannot quite match the pony anime model when it comes to interactions with males. It requires more patience to test different spell combinations to achieve the desired actions. Of course, you're welcome to modify and customize it yourself to better suit your needs. Ultimately, I hope you have an enjoyable experience using it.
It's really odd, but whenever I generate pics, no matter the sampler, at the final steps it's like the image fades out and gets foggy. Like all colour leeches out.
(see samples in the gallery)
Thank you for your feedback. To enhance the vibrancy of the colors, you could utilize VAE to refine the image.
@chatgpt2023126 Whic VAE do you recommend? I used the default SDXL VAE
@tvange365 You could try using this VAE model.
https://civitai.com/models/152040?modelVersionId=360299
Something is not quite right with v2 compared to v1. It seems to ignore part of the prompt, most of the time and the longer the prompt the worse it gets.
For example: "1girl, leaning forward, over box, from side, outdoors, heels, shirt, skirt, pink hair, sunglasses, sun hat, ", with v1 it'll mostly work, but with v2 the box or sunglasses will be missing or she wont be leaning forward etc.
Thank you for your feedback. However, I will continue to use the prompt you provided, as it suits my needs. You could check the examples on the Tensor.Art.
https://tensor.art/images/734409072540988263?post_id=734409072536793960&source_id=nzy2oFHrnUWwoPAtYX_x8BAm
Just did a xyz comparison with a ton of 'realistic" pony models without upscaling.
This model by far looked the most realistic without upscaling. Skin texture, eyes and hair all looked significantly better.
Tradeoff however is that it did not follow the prompt at all...
If the next version get better at following a prompt without losing too much of the realism, then we have a winner
Thank you very much for your feedback and suggestions. I understand the shortcomings of this model, and I hope to create a model next time that better adheres to the prompt while maintaining its realism. Although this may be a challenging task for me, I am committed to improving it.
@chatgpt2023126 Just noticed you posted a newer version so will definitely check it out. The model looks promising at least.
@Asaghon Thank you for your continued interest and attention. This is a beta version, and I hope everyone can help test it out so that I can look forward to integrating an even better version in the future.
This current version seems to follow the prompt fairly well, though it helps to use the "correct" tags for what you want, since this is pony-based:
https://danbooru.donmai.us/wiki_pages/tag_groups
@Gouda Yea was using correct tagging all along :p Directly comparing V2 and V2.5 now, V2 does barely any of the tags while V2.5 does. V2 just seems to want to draw a human and nothing else. (altough the images look great)