Feat/align vid rebase#1824

Open

BossunWang wants to merge 55 commits intokijai:mainfrom

BossunWang:feat/AlignVID_rebase

Commits on Dec 19, 2025

feat:Implement AlignVid ASM with energy-based attention scaling
BossunWang
committed
feat:Enhance AlignVid ASM with energy range parameters and configuration node
BossunWang
committed

Commits on Dec 24, 2025

Support loading FlashPortrait model

kijai
authored and
BossunWang
committed
Add portrait_cfg

kijai
authored and
BossunWang
committed
Update nodes_sampler.py

kijai
authored and
BossunWang
committed
Stand-in RoPE adjustments

kijai
authored and
BossunWang
committed
Possibly fix uni3c + multitalk

kijai
authored and
BossunWang
committed
Support mixed precision fp8 scaled models

kijai
authored and
BossunWang
committed
Squashed commit of the following:
BossunWang
committed
Remove print

kijai
authored and
BossunWang
committed
LongCat-Avatar example

kijai
authored and
BossunWang
committed
Update LongCatAvatar_audio_image_to_video_example_01.json

kijai
authored and
BossunWang
committed
version 1.4.5

kijai
authored and
BossunWang
committed
Allow loading StoryMem -LoRAs

kijai
authored and
BossunWang
committed
fix: Implement AlignVid ASM with energy-based attention scaling
BossunWang
committed
feat: Add AlignVidScheduler for energy-based attention scaling
BossunWang
committed
fix: Add energy range parameters to AlignVid ASM for improved scaling
BossunWang
committed
Merge branch 'main' into feat/AlignVID_rebase
BossunWang
authored

Commits on Dec 26, 2025

fix: Refactor scale_keys usage in AlignVid model for consistency
BossunWang
committed

Commits on Jan 5, 2026

Allow higher lora scale

kijai
authored and
BossunWang
committed
Automatically adjust LoRA alpha for Peft rs_lora weights

kijai
authored and
BossunWang
committed
Add node to add story memory latents

kijai
authored and
BossunWang
committed
StoryMem latents are supposed to be encoded one by one

kijai
authored and
BossunWang
committed
Fix uncond variable name when using zero star or fresca

kijai
authored and
BossunWang
committed
Adjust StoryMem lora scaling

kijai
authored and
BossunWang
committed
Cleanup: Remove flowedit code, restructure some other code

kijai
authored and
BossunWang
committed
Fix s2v

kijai
authored and
BossunWang
committed
Cleanup

kijai
authored and
BossunWang
committed
Update LongCatAvatar_audio_image_to_video_example_01.json

kijai
authored and
BossunWang
committed
Cleanup example_workflows folder some

kijai
authored and
BossunWang
committed
Fix LongCat-Avatar audio padding when not enough audio provided for given window

kijai
authored and
BossunWang
committed
Fix context window ref latent device

kijai
authored and
BossunWang
committed
I don't know why this suddenly errors

kijai
authored and
BossunWang
committed
Fix enhance-a-video

kijai
authored and
BossunWang
committed
Add node for SVI 2.0 Pro

kijai
authored and
BossunWang
committed
Adjust scheduler graph drawing

kijai
authored and
BossunWang
committed
Update nodes.py

kijai
authored and
BossunWang
committed
Fix s2v

kijai
authored and
BossunWang
committed
Add node to set attention mode per step and/or blocks

kijai
authored and
BossunWang
committed
Adjust ultravico frame_tokens

kijai
authored and
BossunWang
committed
remove print

kijai
authored and
BossunWang
committed
Fix er_sde steps

kijai
authored and
BossunWang
committed
Fix res_multistep steps

kijai
authored and
BossunWang
committed
Add node for UltraVico params

kijai
authored and
BossunWang
committed
Update nodes_model_loading.py

kijai
authored and
BossunWang
committed
Correct infinite talk loop log entry

kijai
authored and
BossunWang
committed
Make VAE memory reporting optional to reduce log spam, and other logging updates

kijai
authored and
BossunWang
committed
Update wan_video_vae.py

kijai
authored and
BossunWang
committed
feat:Implement AlignVid ASM with energy-based attention scaling
BossunWang
committed
feat:Enhance AlignVid ASM with energy range parameters and configuration node
BossunWang
committed
Support mixed precision fp8 scaled models

kijai
authored and
BossunWang
committed
Squashed commit of the following:
BossunWang
committed
LongCat-Avatar example

kijai
authored and
BossunWang
committed
fix: Add energy range parameters to AlignVid ASM for improved scaling
BossunWang
committed
feat: Refactor get_scheduler function by removing unused alignvid_scheduler parameter
BossunWang
committed