Workflow - V2V Expand Any Video with IC-Lora Outpaint
V2V Expand Any Video with IC-Lora Outpaint
Expand any video in any direction. Outpaint with new content that fits the input video and your prompt.
You can use it to outpaint on all sides, top, bottom, or any combination — and the model fills in content that is visually and temporally consistent with the original footage. Thanks to the great IC-Lora-Outpaint by @oumoumad in combination with padding nodes by KJNodes from Kijai that pads black area to inpaint.
The workflow can add audio in case of a silent video, generated by LTX, or you can keep the original audio from the input video if it has audio.
Video above are some random ones from CivitAI used as examples. Credits to the original creators.
Feel free to try it out ;-)
To download the lora: https://huggingface.co/oumoumad/LTX-2.3-22b-IC-LoRA-Outpaint
Workflow: https://huggingface.co/RuneXX/LTX-2.3-Workflows/tree/main/Video-2-Video
for darker video there is a simple toggle switch in the workflow to apply a gamma fix, that i forgot to use with the girl and the car, entirely my fault there is a little bit of "shadowing" in that one. ;-)
Btw, did the car have 2 plates? 🤔 it doesn't looks like a reflection 😅
Btw, did the car have 2 plates? 🤔 it doesn't looks like a reflection 😅
Depends on the seed i bet ;-)
Btw, why can't i changed the text projection model? felt like it's locked for that sub-directories 🤔 meanwhile, i can changed the text projection model location on your PromptRelay workflow without this kind of issue.
Thats odd. Try right click that node, and choose "re-create node"
Thats odd. Try right click that node, and choose "re-create node"
Recreate node (keeping the value) doesn't seems to make a difference, but if i choose "Fix node (recreate )" it became like this:
Is the text projection model not supposed to be in models/text_encoders folder? 🤔 since i can't change the value to anything other than undefined.
Oops I wasn't aware that your PromptRelay workflows are using a different node (DualCLIPLoader) for the text projection model 😅 i guess LTXV node expect the file to be somewhere else 🤔
ah yes ;-) I didnt even notice you used the LTX one. Think that looks in checkpoints
While the Dual Clip loader looks in text_encoder folder
Thats probably my bad. Both can be used. But better to use the dual clip loader since i use that in all other workflows.
I'll update ;-)