WAN2.1 | FLUX/SDXL to I2V | Interpolate + Upscale.
Unhappy with (most) WAN T2V workflows decided to combine a (good) I2V flow with a FLUX flow to generate the image fed to I2V (making it effectively a T2V flow).
The I2V part is mostly taken from https://civarchive.com/models/1309369 from UmeAiRT (with some slight modifications).
An image chooser between the FLUX generated image allows you to select one of the images that would be fed to the rest of the flow.
Both GGUF and Native models can be used (tested with the Q5/Q8 GGUF and KijaWan's 14B 480p models).
Update: a 2nd version using SDXL has been added.
Description
WAN2.1 | FLUX/SDXL to I2V | Interpolate + Upscale.
Unhappy with (most) WAN T2V workflows decided to combine a (good) I2V flow with a FLUX flow to generate the image fed to I2V (making it effectively a T2V flow).
The I2V part is mostly taken from https://civitai.com/models/1309369 from UmeAiRT (with some slight modifications).
An image chooser between the FLUX generated image allows you to select one of the images that would be fed to the rest of the flow.
Both GGUF and Native models can be used (tested with the Q5/Q8 GGUF and KijaWan's 14B 480p models).
Update: a 2nd version using SDXL has been added.
FAQ
Comments (4)
SDXL -> WAN
Looks amazing! Do I need to trigger something specific when choosing a image or is clicking on it enough? because nothing happens that way for me. No errors when opening the workflow
The usual way that the "Image Chooser" would work (even if you set it to one single image), is that you select the one to process followed by "Process Selected image". It would then send that one to the rest of the WAN I2V workflow part.
what does this do? using flux/sdxl models/loras with WAN?
It would use a SDXL (or Flux) model to generate an initial image. That generated image is then used in a I2V WAN flow. The benefit is also, to be able to use WAN I2V lora's while still having a 'T2V' based flow.