Workflow Article: Kytra's Cosmic Garden - Animated AI Landscape Generator | Civitai
VIEW MORE EXAMPLE OUTPUTS HERE AT MY MODEL LINK:
The examples shown for this workflow all used this aesthetic model:
V3_SD15_Adapter_Lora is also loaded into the 'LoraLoaderModelOnly'
The examples shown for this workflow all used this Motion Model:
https://civarchive.com/models/326698/animatediff-lcm-motion-model
You should only need to drop an IP adapter image into this Load Image node to drive the composition of the landscape.
I've included the nonsense positive prompt prefix that I used in my examples.
Feel free to change it, but the idea is that you provide a prefix of 'image qualifiers' and then WDTagger provides generalizations on the image.
The controlnet should be loading sd15 tile controlnet.
If you don't change any of the resolution settings, this workflow should complete one render in ~4 minutes and consume 19GB of VRAM (I wasn't optimizing for low VRAM usage)
Description
FAQ
Comments (3)
Error occurred when executing IPAdapterTiled: Missing CLIPVision model.
The clip vision model might have a different name than the one I used in my workflow. Maximize the few model loaders I collapsed in ComfyUI and make sure they're pointing to the appropriate type of model.
The error you're receiving is very common and indicates we have named our models differently from yours.
Error occurred when executing IPAdapterTiled: Error(s) in loading state_dict for Resampler: size mismatch for proj_in.weight: copying a param with shape torch.Size([768, 1280]) from checkpoint, the shape in current model is torch.Size([768, 1024]).