Missing OmniNFT_converted_lora.safetensors causes fallback to T2V (ignores reference image)
Hi, I'm trying to run it on Runninghub (a cloud ComfyUI platform), but I ran into an issue where the generation completely ignores my input reference image. It essentially falls back to pure text-to-video (T2V), making the reference image useless.
After some troubleshooting, it seems the issue is caused by the missing OmniNFT_converted_lora.safetensors in the Power Lora Loader (rgthree) node. Because the platform lacks this specific LoRA, the node fails to load properly, which breaks the Image-to-Video (I2V) conditioning chain downstream.
Could you please point me to where I can download this specific OmniNFT LoRA?
I managed to find and use Kijai's LTX-2.3-OmniNFT-RL-Lora_bf16.safetensors as a replacement for the missing LoRA. but the output is still purely Text-to-Video. The reference image is completely ignored in the final generation.
Uh it's linked in the nodes on the left side of the workflow.
I managed to find and use Kijai's LTX-2.3-OmniNFT-RL-Lora_bf16.safetensors as a replacement for the missing LoRA. but the output is still purely Text-to-Video. The reference image is completely ignored in the final generation.
That means the i2v node was accidentally turned to bypass, check that.