Pre-release
barely works- use "crouching position" and "sliding". Character must already be in this position. Trained for 4 directions with 121 frames
rank32
resolution 512
4-bit text encoder
AIToolKit
The pose is captioned 1 to 1 (basic grammar)
LTX2-GoldenSlide-512-4bitTextEncoder-Rank32_000001000.safetensors