(HUNYUAN v1) Trigger word "Valkyrae". Drag and drop a video for the workflow I used. Follow THIS GUIDE to train your own Hunyuan video LoRA on Windows. Love my models and want to support me? Send some buzz my way. If you're feeling extra generous and would like to donate to help support the costs of training these resource intensive models you can directly donate to me though HERE.
(FLUX) Just prompt Valkyrae. For best results use the full Flux1.D model and fp16 clip. FP8 model works as well. Trying 4,000 steps on this one. Love my models and want to support me? Send some buzz my way. If you're feeling extra generous and would like to donate to help support the costs of training these resource intensive models you can directly donate to me though here
Description
FAQ
Comments (5)
Amazing work as always! You've got the process locked in!
I ran into some snags getting diffusion-pipe working in the cloud with its additional layers of complexity so I sent an e-mail hoping to get that up and running. Sprinkle a lucky wish on that if you can.
Did you train 512x512 or 1024x1024? Did you train wíth video clips or only on images? Thanks.
I've done both 1024 and 768. Didn't see a decrease in quality with 768 so that's what I've stuck with. All made with only images.
@AIENGI Okay thanks! Did you train locally with a 4090 or on some remote runtime?
@steffangund 4090 locally on WSL with diffusion pipline.