Search Results
7/20/2025, 5:58:05 PM
>>8665648
There are LORA training guides in the OP. Other than those you can search online and there are plenty. One I keep referring back to is https://civitai.com/articles/138/making-a-lora-is-like-baking-a-cake
>On that note, what's the bare minimum quality for an image to be worth including in the training data? I don't have all that much to work with.
Quality matters more than quantity for LORA training, so you want the best quality images you can get especially if there are only a few. You might want to carefully upscale what images you do have if they're low-res. You can also cut up an image if it's like a character profile and use each individual image to get a larger dataset. And you can double your dataset with flip augmentation too.
Pic related is an old table about LORA training with recommended amount of images and settings and stuff. It is old though so take it with a grain of salt.
There are LORA training guides in the OP. Other than those you can search online and there are plenty. One I keep referring back to is https://civitai.com/articles/138/making-a-lora-is-like-baking-a-cake
>On that note, what's the bare minimum quality for an image to be worth including in the training data? I don't have all that much to work with.
Quality matters more than quantity for LORA training, so you want the best quality images you can get especially if there are only a few. You might want to carefully upscale what images you do have if they're low-res. You can also cut up an image if it's like a character profile and use each individual image to get a larger dataset. And you can double your dataset with flip augmentation too.
Pic related is an old table about LORA training with recommended amount of images and settings and stuff. It is old though so take it with a grain of salt.
Page 1