thanks for the buzz 🤗

@Tyberzandari when i was gather training data, it seemed like there was enough for BT, most of the art was biased towards him so this lora ends up making him randomly all the time. there might not be enough diverse and high enough resolution to create a lora for some of the other specific Titans

@Tyberzandari there's issues when there isn't enough diversity in the dataset, I mostly used fan art for this one, and used in game screenshots very sparingly. There are also issues when most of the dataset is 3d renders. I had some models come out really badly when that dataset doesn't have enough variety that I haven't published. I know some people know how to do things with 3d models and get different angles to work into the dataset, but that's not something I know how to do. However, I'm not an expert at illustrious training, I only started doing it recently, so maybe there are some people that would be able to do it well. I might come back to it if if I get better at this and think I can pull it off.

@Tyberzandari np, the way to get through it as far as I know is to do multiple training runs and to generate as good images as you can that are in a different art style / context with the last one until you have a good enough dataset you made yourself with the failed runs. That's the way I know how to brute force through not having enough training data, but it's very time and resource consuming, but if it's something that you really want to make, there is a path forward.

thank you, I don't really know about gundam, I'll have to learn about it more first, maybe do some specific models first

@7117 I looked, there seems to be a lot of gundam loras for specific mechas, if you search "gundam", it's about halfway down until the mechas start appearing, did you not see these before or are these loras lacking something? and thank you