Initial comfy setup
For v1.0
1) Weights to generate
2) Github the main node
For v2.0
1) I2V Quant Model -> models/diffusion_models
2) CLIP VISION -> models/clip_vision
3) VAE -> models/vae
Prompt can write Florence2 LLM (workflow has toggle between Florence2 and manual user prompt).
Tested on 4080 16GB
v2.0 dev in development. If you have problems with AnyPython node, you should switch the channel recent or/and lower security_level (normal- OR weak) in the ComfyUI/Custom_nodes/ComfyUI-Manager/config.json
