 kijai
|
b9e49af4d1
|
Just resize any dimension mismatch on uni3c input
|
2025-09-10 17:18:02 +03:00 |
|
|
kijai
|
c254ae4235
|
Rather interpolate any Uni3C input frame count differences
|
2025-09-10 16:47:13 +03:00 |
|
|
kijai
|
dd36ab46c6
|
Pad Uni3c input when using VACE ref
|
2025-09-10 16:26:20 +03:00 |
|
|
kijai
|
7cb46ca8c3
|
Allow Uni3C to work with T2V
|
2025-09-10 16:09:44 +03:00 |
|
|
kijai
|
e5ef9752a7
|
Fix fp8_fast when not using loras
|
2025-09-08 18:54:29 +03:00 |
|
|
kijai
|
fe379fa77f
|
InfiniteTalk: Add option to save results during the process, make it clearer no latents are returned
|
2025-09-08 18:50:53 +03:00 |
|
|
kijai
|
e5955d8395
|
InfiniteTalk: Don't create 2 windows if total frames requested is under frame_window_size
|
2025-09-06 23:10:26 +03:00 |
|
|
kijai
|
011c0ce38d
|
InfiniteTalk: Pad with silent embeds instead of repeat, add MultiTalkSilentEmbeds -node
|
2025-09-06 21:58:03 +03:00 |
|
|
kijai
|
ae768f53a4
|
Make merge_lora switches with multiple loras behave like it used to and how the tooltip indicates
|
2025-09-06 19:28:16 +03:00 |
|
|
kijai
|
bafd5503ac
|
Fix unianim error when using ref pose
|
2025-09-05 23:38:08 +03:00 |
|
|
kijai
|
edea93286c
|
Small potential tweaks to Pusa vram use
|
2025-09-05 16:00:18 +03:00 |
|
|
kijai
|
fbf8f169b6
|
Update nodes.py
|
2025-09-05 15:59:01 +03:00 |
|
|
kijai
|
0ad2b00cc0
|
Fix force_offload on Multi/InfiniteTalk long I2V node
|
2025-09-05 15:54:11 +03:00 |
|
|
kijai
|
9b864b987b
|
Don't error even if redundant offloading is enabled for single text encode
|
2025-09-04 16:20:44 +03:00 |
|
|
kijai
|
cbbf44f3bc
|
Add Pusa 2.2 example
|
2025-09-04 02:22:54 +03:00 |
|
|
kijai
|
c38da2242d
|
version
|
2025-09-03 22:49:02 +03:00 |
|
|
kijai
|
d05796ffb3
|
Update requirements.txt
|
2025-09-03 22:48:22 +03:00 |
|
|
kijai
|
488b6628c7
|
typo
|
2025-09-03 22:47:42 +03:00 |
|
|
kijai
|
8131eaef23
|
update tooltip
|
2025-09-03 22:26:07 +03:00 |
|
|
kijai
|
242637db5c
|
Accept per latent noise multiplier list for Pusa as well
|
2025-09-03 22:18:37 +03:00 |
|
|
kijai
|
24ec3f6960
|
Add WanVideoLoraSelectByName
select lora by name string
|
2025-09-03 20:22:41 +03:00 |
|
|
kijai
|
02da8fed47
|
sigma plot tweaks
|
2025-09-03 20:01:55 +03:00 |
|
|
kijai
|
7a29bcb18a
|
Fix sigma graph
|
2025-09-03 19:49:00 +03:00 |
|
|
kijai
|
617ce55938
|
Fix up pusa noise multipliers indexing
|
2025-09-03 19:24:47 +03:00 |
|
|
kijai
|
e76bf4d899
|
Add Pusa additional noise handling
|
2025-09-03 18:32:34 +03:00 |
|
|
kijai
|
493106b555
|
Revert this as it didn't end up working
Pusa extra latents seem to need to be encoded individually
|
2025-09-03 15:48:14 +03:00 |
|
|
kijai
|
54391300b1
|
Never use main_device with unmerged loras
|
2025-09-03 10:00:51 +03:00 |
|
|
kijai
|
c57ffd8150
|
update example
|
2025-09-03 00:41:23 +03:00 |
|
|
kijai
|
838e095dbd
|
Update model.py
|
2025-09-02 23:46:04 +03:00 |
|
|
kijai
|
863ea45d23
|
Update model.py
|
2025-09-02 23:34:34 +03:00 |
|
|
kijai
|
44f9584f16
|
Fix S2V
|
2025-09-02 23:31:47 +03:00 |
|
|
kijai
|
cc3b79ab18
|
Update nodes.py
|
2025-09-02 21:08:02 +03:00 |
|
|
kijai
|
cdfa3a8661
|
Update nodes.py
|
2025-09-02 21:05:04 +03:00 |
|
|
kijai
|
1492310544
|
Update nodes.py
|
2025-09-02 20:49:24 +03:00 |
|
|
kijai
|
0ad24cd7e7
|
Allow for multiple I2V images for context windows
Similarly as multiple prompts already work, use the new WanVideoEncodeLatentBatch -node to encode them
|
2025-09-02 19:28:33 +03:00 |
|
|
kijai
|
c538cd4c8a
|
Fix unmerged lora
oops
|
2025-09-02 18:48:25 +03:00 |
|
|
kijai
|
3183589eff
|
Merge branch 'main' of https://github.com/kijai/ComfyUI-WanVideoWrapper
|
2025-09-02 17:09:12 +03:00 |
|
|
kijai
|
4cf23a5a40
|
Can't do this inplace with all models
|
2025-09-02 17:09:05 +03:00 |
|
|
Jukka Seppänen
|
0405dde5a2
|
Merge pull request #1165 from zinigor/fix/unbound-local-error
Moved dimension update to where the count is initialized.
|
2025-09-02 15:06:44 +03:00 |
|
|
Igor Zinovyev
|
42054724b7
|
Moved dimension update to where the count is initialized.
|
2025-09-02 14:32:01 +03:00 |
|
|
kijai
|
42d7180759
|
keep scale_weight on gpu to avoid needless casts
|
2025-09-02 01:04:16 +03:00 |
|
|
kijai
|
82608009c8
|
use a copy of the scheduler if provided by the scheduler node to make sure it's always reset
|
2025-09-01 23:04:05 +03:00 |
|
|
kijai
|
f9755820de
|
handle negative latent indices with extra latents
|
2025-09-01 14:19:51 +03:00 |
|
|
kijai
|
56e3366c26
|
Fix pusa steps
|
2025-09-01 13:50:41 +03:00 |
|
|
kijai
|
43d8a2c8a5
|
flowmatch_pusa split sampling fixes
|
2025-09-01 13:34:49 +03:00 |
|
|
kijai
|
d4d942c5ee
|
Update nodes.py
|
2025-09-01 11:17:22 +03:00 |
|
|
kijai
|
30c2fca799
|
WanVideoScheduler fixes
support flowmatch pusa when using the node
|
2025-09-01 11:07:41 +03:00 |
|
|
kijai
|
d9038d575f
|
Allow setting both start and end step by sigma
|
2025-08-31 19:04:01 +03:00 |
|
|
kijai
|
b8acd3ee48
|
Allow using fp16 multitalk model with GGUF again
|
2025-08-31 18:57:59 +03:00 |
|
|
kijai
|
9f84cd2254
|
Fix offloading when using stand-in lora
|
2025-08-31 18:45:13 +03:00 |
|