
Odinson-SDXL
+ vid2vid
+ OpenPoseXL2
+ animateDiff XL
+ Lora loader
(default strength 1.0 <- change this)
Versions:
-
V9 - initial release (used OpenposeXL by default, slow)
-
V10 - desaturated input (used with Lora Sketch XL, fast)
-
V11 - same as V10, (Interpolation now after ksampler, fast)
use comfyui manage to install missing nodes & models
you will need:
Any SDXL Checkpoint
Any SDXL Lora
Any SDXL ControlNets
Rife49.pth (comfyui frame interpolation)
ip-adapter_sdxl_vit-h.safetensors (comfyui ipadapter plus)
CLIP-VIT-H-14-laion2B-s32B-b79k.safetensors (CLIP Vision)
mm_sdxl_v10_beta.ckpt (animateDiff evolved)
check the screenshots below!
at 1024 expect ~22GB VRAM
at 768 expect ~12GB
reduce the dimension to reduce the load.
Be careful with your video size ! this is 1024x1024 (large)
-
Pick video input and set the frame range:
-
Check the maximum size for the longest side (clamp)
-
Check Frame interpolation (2 = double input frames)
-
select the ControlNet (OpenposeXL2 shown)
-
Check IP Adapter settings and load style image
-
Set final resolution here (use 1024 for no upscale)
-
Set Final Framerate
-
Load Lora & check Animatediff settings (should be the same)
-
Finally set your Prompt and you can start the generation
描述:
defaults to Sketch, mucht faster.
depth and openpose estimation can triple the prep time.
训练词语:
名称: odinsonSDXL_v10.zip
大小 (KB): 5
类型: Archive
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success