🎉 My first original video Lora is here! Come study together 🧐
Guys, after 13 hours and using 4 RTX 4090 graphics cards running simultaneously, I finally completed the first original trained video Lora 👏.
To find the most stable version, I tested 5 Loras with different epochs, and this current version is super stable 🌟. I've uploaded it to lib because I want to learn, research, and test it together with everyone 🤝.
I truly welcome your feedback! If you get good results, feel free to share screenshots 📷. Those who share three screenshots can message me to get the unencrypted version, which allows local operation 😎.
Since this is my first time training a video Lora, there might be some imperfections. I'm still exploring many parameters and techniques, so please bear with me 🥺. Trolls, please stay away—they're not welcome here 🙅.
The best part about lib is that it supports direct model invocation to generate videos, so even 小伙伴们 who don't know workflows can get started easily 👏.
Here are some recommended parameters:
Weight: 0.8–1 (adjust for different images).
Base model: This Lora was trained with wan14B-480p, so using this model is recommended.
Generation time: 5 seconds.
Workflow: Coming soon!
Prompt: A person suddenly appears and kisses the protagonist in the video
Others: Will be updated based on feedback.
Go give it a try, guys! Can't wait to see your results～
#OriginalVideoLora #DeepLearning #VideoGeneration #LoraTraining #AICreation