Instructions to use Alissonerdx/LTX-Best-Face-ID with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Inference
- Notebooks
- Google Colab
- Kaggle
Movements somehow
Thank you for your hard work! You have posted new examples using the CharacterSheet. I didn't quite understand the dancing examples. Did you add your movements somehow? for example, here: https://huggingface.co/Alissonerdx/LTX-Best-Face-ID/blob/main/examples/sample_9.mp4
They are using an OpenPose style (coloured stick figure) control video to control the character movement, rather than directly using a video of someone dancing, for example.
This helps to avoid elements from the control video bleeding into the final output.
If you are using ComfyUI, you can look in Templates/Utility, and there is a workflow there called 'SDPose-OOD: Video to Pose Map', which allows you to input a video of someone dancing, etc, and have it output an OpenPose control video.
You can combine this IC LoRA with other IC LoRAs , one of them is https://huggingface.co/Lightricks/LTX-2.3-22b-IC-LoRA-Union-Control. It allows you to send pose, canny, and depth maps. In my case, I extracted the pose from a video and used their IC LoRA along with it. I'll try to upload a basic workflow with this, then you just need to adapt it.
Workflow with Union Control
Thanks! I'll try. I made a video tutorial based on your models. Unfortunately, I don't speak English.. :https://youtu.be/WHq2O4gxdxM?si=FuH4T_afhq-xEFa5
Workflow with Union Control
Thanks! I'll try. I made a video tutorial based on your models. Unfortunately, I don't speak English.. :https://youtu.be/WHq2O4gxdxM?si=FuH4T_afhq-xEFa5
Hahaha, don't worry, just posting in your own language already helps a lot. I don't speak fluent English either; my native language is Brazilian Portuguese. I'll watch your video when I get home.