even does text. loads way faster than old qwen models

There's even an anime model.

https://civitai.com/models/2259646?modelVersionId=2543657

The FP16 version is a little better, but its not normally noticeable on smaller images.

Qwen model is even more impressive, however it sometimes overdetails things.

Look what came today

Now she's with me every day! ❤️😭

The new Dasiwa workflows and SynthSeduction models on Civit have finally resolved the issue for me of the loops it makes having discoloration on the last frame. I feel like its worth it to make videos now. I wonder if I can add sound?

https://civitai.com/models/1981116

Trying different quantizations of the model to see how it effects speed and quality.

I have to say the larger models do a much better job.

The unquantized model does very well, but also sometimes still moves the question marks with her head.

You can follow me on twitter, I'll post some art there

I can't follow you back rn due to daily follow limits or something, but I'll get around to it if I followed you before.