When using the <extend> feature, is there a way to apply LoRAs to specific <extend> sections?

[Lainier]

When I add LoRAs to extended videos, they apply to the whole video. I want to use LoRAs to specific sections. Can it be done? I suppose not, but it would be a great feature :) By the way, I'm using WAN 2.2.

Here's an example. Suppose I have the LoRA xyz and the LoRA abc and I want to extend a 5 second video to a total of 15 seconds. My prompt would be:

base video prompt text here <extend:81> extended video prompt text here <extend:81> extended video prompt text here <video> <lora:xyz_high_noise:1> <lora:abc_high_noise:1> <videoswap> <lora:xyz_low_noise:1> <lora:abc_low_noise:1>

When I use this kind of structure, xyz and abc LoRAs are applied to the whole video, but I want to apply each to a different extend section. For example, xyz to the first <extend> prompt and abc to the second <extend> prompt.

[brendanhoar]

In other tag domains (segments? regions?) I have found that you can add a lora back in with a negative weight to simulate "removing" it from the in-memory base model. Try something like the following to use the lora xyz in the first set of frames and lora abc in the second set:

base video prompt text here<lora:xyz:1.2>
<extend:81>
extended video prompt text here <lora:xyz:-1.2><lora:abc:0.9>

[Lainier]

In other tag domains (segments? regions?)

I forgot to mention I'm using WAN 2.2, so there's further complication. I've updated my post to better explain myself.