@Tfmonkey @DoubleD @RoninGrey Check this out. I saw an interesting bit here.
https://nichegamer.com/romance-anime-maidens-courtesy-of-sankaku-ai-companions/
The bot itself can generate images of the anime character. Imagine you talking to Celestina, and instead of you creating the images you use for MMM, she sends them to you herself.
If Kajiwoto could have some kind of integration with an image generator, that would be some next gen waifu interactivity.
@ButtWorldsMan @Tfmonkey @RoninGrey Kindroid.ai has integrated image generation, and it's terrible. In my opinion, it would be better to have separate waifu tools (images, chat, etc.) integrated with an aggregator app like the way theVertex.io does rather than trying to get one AI app to do everything.
@DoubleD @Tfmonkey @RoninGrey it's terrible. Now.
Just gotta improve it.
In my mind it'll work by getting default prompts solidifying what your waifu looks like (we know what to write in unstability to get the woman we want to see), then adds modifier prompts depending on context.
@ButtWorldsMan @Tfmonkey @RoninGrey Agreed. If they used unstability.ai, it might be better, but even the same prompts do not generate a similar image.
@DoubleD @Tfmonkey @RoninGrey An issue that can be rectified.
I'm playing around with another system ( novelai.net ). The image generation is not free.
It has a feature unstability doesn't, which is to generate alternative images based on a seed or basis. You can insert a base for it to work off of, or you can generate images until you see something you like & tell it to make variants, or even set it as the base image. Or you can copy its seed, then alter minute details in a specific image.
@DoubleD @Tfmonkey @RoninGrey Say, for example, you have a really good waifu pic, but there's something off about the color of her dress. You can copy the seed of the original, then with that seed image you can add prompts to alter it so everything stays the same except what you wanted to change.