This workflow has 20 presets set up with lora loading in preset text boxes.
You can set it to run just 1 preset or set it to randomize infinitely.
you can set ref_image_0 or 1 to load from folders and just need to change 1 number for the total amount of images to have it randomly select 1 image from the folder on every run, or just set as a single reference with a switch!
ref images in ref2va have been tested with
ref_0"female"
ref_1"male"
ref_2"room/background"
Updated for r2va and fl2v+audio
The loras used in the workflow are listed in a note on the main work page.
Repos for custom nodes used in these wokflows incase not using comfy-manager
https://github.com/rgthree/rgthree-comfy
https://github.com/melMass/comfy_mtb
https://github.com/kijai/ComfyUI-KJNodes
https://github.com/yolain/ComfyUI-Easy-Use
https://github.com/nova452/Rebalance-Pack
https://github.com/TensorVizion/Minimax-H3-Video-Music-Nodes
https://github.com/cubiq/ComfyUI_essentials
https://github.com/wallen0322/ComfyUI-Wan22FMLF
https://github.com/ltdrdata/was-node-suite-comfyui
https://github.com/ltdrdata/ComfyUI-Impact-Pack
Description
Tested side profile and pov style images. Converted ref2va to a new flow for FL2V+Audio, all model cards ran on a rtx3060 w/64gb ram turbo4step lora at 8x steps.