After much trial and error, I finally got my first Hunyuan Lora up and running! It's not perfect by any means as this is my first time training on Hunyuan so expect a lot of misses. I still have a lot to learn so expect a V2!
"top_off, shirt_lift" are the triggers words I usually put them at the front and end of the prompt, but you might be able to get away with just one or the other. Also try using "She is lifting her (DESCRIBE TOP HERE) with her arms crossed, then lifts the (SAME TOP FROM BEFORE) above her head with her arms up, revealing her (BREASTS, BRA, ETC.)"
You can also use my prompts in each video as I will be uploading each prompt there. The provided videos were made at "288 by 480" resolution at 24fps then upscaled 4X and at 60 fps using Topaz labs. One video was at native "360 by 648" resolution.
The dataset consisted of twenty videos rounding to about 40 - 100 frames each. The videos were downsized to "288 by 480" resolution. Here are my settings that worked for me below. Honestly, I'm still new to this so if anyone finds the "Secret Sauce" for getting the right settings please let me know and I'll change accordingly. Thanks!
Trigger word: top_off, shirt_lift
Strength: 1 - 1.3
Shift: 7
Guidance: 7
Steps: 20
FPS: 24
Frames: 49 - 85 (more could work but wasn't trained too much after around 100 frames)
V1 (Hunyuan)
After much trial and error, I finally got my first Hunyuan Lora up and running! It's not perfect by any means as this is my first time training on Hunyuan so expect a lot of misses. I still have a lot to learn so expect a V2!
"top_off, shirt_lift" are the triggers words I usually put them at the front and end of the prompt, but you might be able to get away with just one or the other. Also try using "She is lifting her (DESCRIBE TOP HERE) with her arms crossed, then lifts the (SAME TOP FROM BEFORE) above her head with her arms up, revealing her (BREASTS, BRA, ETC.)"
You can also use my prompts in each video as I will be uploading each prompt there. The provided videos were made at "288 by 480" resolution at 24fps then upscaled 4X and at 60 fps using Topaz labs. One video was at native "360 by 648" resolution.
The dataset consisted of twenty videos rounding to about 40 - 100 frames each. The videos were downsized to "288 by 480" resolution. Here are my settings that worked for me below. Honestly, I'm still new to this so if anyone finds the "Secret Sauce" for getting the right settings please let me know and I'll change accordingly. Thanks!
Trigger word: top_off, shirt_lift
Strength: 1 - 1.3
Shift: 7
Guidance: 7
Steps: 20
FPS: 24
Frames: 49 - 85 (more could work but wasn't trained too much after around 100 frames)