Tensor.Art
Create

The Araminta Experiment

CHECKPOINT
Reprint

Updated:

Fv5
Fv4
Fv3
Gv2
Fv2
Ev3
Fv1

No showcase images available, the model won't be visible to others.

Now

Comparison gallery here: Ev4 - Ev3 and Fv1 - Ev4

Current SOTA model in my experiment:

Base model: Fv1 or Ev4 (if you accept unexpected NSFW) or Cv6... Fv1/Ev4 are biased toward NSFW, but still better in many aspects.

NSFW model: Fv1

Illustration : Fv1 (NSFW) or Cv6 (SFW)

Starting from E serie, models are evolving sometimes by merging with other models (thanks to other contributors!), but mostly via training on my own dataset: a modest dataset (~2000 images currently), but I try to somehow compensate with quality and originality.

Starting with Fv1, I have included many synthetic images I created using previous versions: playing hard with prompt and retouching when necessary the result in Photoshop in order to have a dataset which contains many original images.

Version Detail

SDXL 1.0
Based on Ev2 and merged with a Lora I trained on about 800 hand picked, high quality images, carefully retouched (think watermark) and manually captioned. And the result is beyond my expectation and validate this approach for further improvement. I will now concentrate on improving my dataset :) This version is basically again a step forward compared with Ev2 with almost everything quite noticeably better: Better NSFW (still room for improvement though) Better illustration styles (will improve it further in the future because I like that) Better faces and skin texture Follow better the prompt Ev3 behave a bit differently thanEv2 though and parameters or maybe prompt may have to be tweaked a bit, but it is because it has more details natively and is better at understanding the prompt.

Project Permissions

Model reprinted from : https://civitai.com/models/463163?modelVersionId=716653

Reprinted models are for communication and learning purposes only, not for commercial use. Original authors can contact us to transfer the models through our Discord channel --- #claim-models.

Related Posts

Describe the image you want to generate, then press Enter to send.