Add native Qwen3-VL vision weights for TI2V

Remove inaccurate model card claims

Publish FastVideo LingBot-Video MoE 30B-A3B conversion

initial commit