kijai
acb662b5af
Merge branch 'main' into bindweave
2025-11-11 11:44:26 +02:00
kijai
907c9e1cdd
Update nodes.py
2025-11-10 21:02:58 +02:00
kijai
68392684b5
Add node to use UniLumos
...
Simply allows fore and background latent inputs for UniLumos relight model, example inputs seem to work: https://github.com/alibaba-damo-academy/Lumos-Custom/tree/main/UniLumos/UniLumos/examples
2025-11-10 19:06:45 +02:00
kijai
a3b2f67337
Pad clip vision embeds like in original code
2025-11-08 16:03:00 +02:00
kijai
1e00c8fb28
Update nodes.py
2025-11-08 12:21:11 +02:00
kijai
f972b31bf2
Update nodes.py
2025-11-07 01:09:22 +02:00
kijai
7bf99791ad
Update nodes.py
2025-11-07 00:44:13 +02:00
kijai
7a5587b5af
Let the user resize for QwenVL
...
Seems to need smaller resolutions
2025-11-07 00:39:10 +02:00
kijai
d6cf172846
Update nodes.py
2025-11-06 23:41:24 +02:00
kijai
8992c6af64
Don't include padding for scheduler
2025-11-06 19:22:56 +02:00
kijai
e4084a961b
Update nodes.py
2025-11-06 18:32:53 +02:00
kijai
3ec1edefbe
init
...
For testing, no idea if it works yet
2025-11-06 17:35:48 +02:00
kijai
64c195167b
Update nodes.py
2025-10-30 17:33:42 +02:00
kijai
d74cfc54e8
Don't zero the extra frames
...
Possible fix for SVI "shot" method
2025-10-24 15:28:35 +03:00
kijai
001040abc7
Update nodes.py
2025-10-21 20:27:37 +03:00
kijai
746d46e137
WanVideoScheduler: Always show split step sigma
2025-10-21 20:14:54 +03:00
kijai
e221fdd7ac
Fix normal tiny vae
2025-10-16 14:54:00 +03:00
kijai
bb75cddd60
Add minimal FlashVSR upscale support
...
https://zhuang2002.github.io/FlashVSR/
This only implements the projection model and the VAE, which seems to be enough for upscaling. This does NOT implement any of the streaming and sparse attention code.
2025-10-15 18:24:28 +03:00
kijai
6c38f8cb24
WanAnimate: Force dimensions to be divisible by 16
2025-10-13 17:31:54 +03:00
kijai
05232fb966
Update nodes.py
2025-10-05 01:37:12 +03:00
kijai
00a5cb8e5c
Update nodes.py
2025-10-05 01:27:32 +03:00
kijai
d380c7d219
Add TSR (Temporal Score Rescaling) to experimental args
...
https://github.com/temporalscorerescaling/TSR
2025-10-05 01:14:25 +03:00
kijai
697fff7442
Add FlowMatchSAODEStableScheduler
...
source: https://github.com/eddyhhlure1Eddy/ode-ComfyUI-WanVideoWrapper
2025-10-02 00:53:49 +03:00
kijai
1c7e32d8af
Support lucy edit
2025-09-22 15:34:48 +03:00
kijai
95cd0ef690
WanAnimate: Use empty frames for face when none are provided and masking is used
...
Necessary for the mask to work properly in this scenario
2025-09-22 15:18:17 +03:00
kijai
02583db554
WanAnimate: Encode pose latents in the loop too for better sync and seams between windows
2025-09-22 02:36:17 +03:00
kijai
6cb1241d48
Refactor: Move sampler code to it's own file
2025-09-21 13:43:18 +03:00
kijai
a6dc8aeb64
offload before vae encode too in WanAnimate loop
2025-09-20 21:52:37 +03:00
kijai
b9fd93c19e
WanAnimate: Fix first latent issue when not using bg_images and looping
2025-09-20 20:14:39 +03:00
kijai
a90ee9773f
Fix VACE first latent auto trimming
2025-09-20 14:51:15 +03:00
kijai
1cc8082452
Fix InfiniteTalk offloading
2025-09-20 14:46:35 +03:00
kijai
0dc8a84564
WanAnimate: zero face pixels for uncond
2025-09-20 01:03:13 +03:00
kijai
795a74d756
Update nodes.py
2025-09-19 22:28:27 +03:00
kijai
14a1f8bd71
Update nodes.py
2025-09-19 21:56:33 +03:00
kijai
de4fcd6fb3
Update nodes.py
2025-09-19 18:40:49 +03:00
kijai
bca7236c73
Update nodes.py
2025-09-19 18:03:24 +03:00
kijai
e4627466f5
WanAnimate block swap fix
2025-09-19 16:59:57 +03:00
kijai
7aa9086b48
WanAnimate bugfixes
2025-09-19 16:23:16 +03:00
kijai
72f5423d6a
Allow uni3c with WanAnimate
2025-09-19 15:28:01 +03:00
kijai
3acfb513ea
WanAnimate input frame count tweaks
2025-09-19 14:51:19 +03:00
kijai
543df461ff
Allow not using face
2025-09-19 13:57:39 +03:00
kijai
b2c8cf969f
Fix pose and face strength setting
2025-09-19 12:04:07 +03:00
kijai
79583d4be3
Support WanAnimate
2025-09-19 10:14:08 +03:00
kijai
95f3307d5c
Fix InfiniteTalk regression
2025-09-16 18:39:04 +03:00
kijai
e9672e8d56
Support HuMo 1.7B
2025-09-16 17:41:07 +03:00
kijai
076c4a5c7e
Experimental: Allow HuMo to work with InfiniteTalk
...
Doesn't work that great, but pushing this anyway for possible future usecases
2025-09-15 19:07:34 +03:00
kijai
7fa6be05b8
Fix HuMo context windows when using cfg
2025-09-13 20:41:16 +03:00
kijai
f6fc9cee6e
Update nodes.py
2025-09-13 20:29:05 +03:00
kijai
699bad3887
InfiniteTalk: Don't stack or return frames when using output path
2025-09-13 20:10:13 +03:00
kijai
9063b99bd1
Allow T2V with HuMo
2025-09-13 19:20:19 +03:00
kijai