Thanks for the great work!
After Stage-1 adaptation (moving from the base model's longer window to the short ~33-frame chunk), did you observe any background/camera jitter or shaking in the outputs? Did it happen for FlashTalk (14B) or FlashHead (1.3B)?
Thanks for the great work!
After Stage-1 adaptation (moving from the base model's longer window to the short ~33-frame chunk), did you observe any background/camera jitter or shaking in the outputs? Did it happen for FlashTalk (14B) or FlashHead (1.3B)?