A workflow uses the Smooth Mix and ComfyUI-PainterLongVideo node to generate a long video with ease. Big thanks to the authors.
All the model details are within the workflow.
Try it online for free: https://www.runninghub.ai/post/1991738572403470338
Please check the video guide before you start using this workflow:
Description
FAQ
Comments (11)
Absolute stunner! Add interpolation and upscale and it will be the best workflow here at Civitai!
If you load the video by coolcoldice into Comfy, they look to have added this to the workflow.
where can i add loras? for example i want to make a dancing video based on a belly dancer loras for the first few seconds, afterwards i want to make the next 5 seconds about her turning back and walking away (bouncy walk loras). So i want to use all those loras separately for each segment of the video. Can this workflow do this?
I'm no expert and I haven't tested this, but I think you would add a 'LoraLoaderModelOnly' between the Get_Model_High and Get_Model_Low and their KSampler in each of the Shot groups. Then each prompt could contain the trigger words for the Lora you are wanting to use at each section.
Edit: gave it a quick test and seemed to work fine.
In Shot 2 onwards are the LoadCLIP Vision and Clip Vision Encode nodes required? The workflow when downloaded has them disabled in Shot 2 and Shot 3 and everything else enabled. When you enable the other shots, they become enabled. Wondering if they need to be disabled?
works on comfy portable, only had to turn off the speed boost section, and add Lora. thanks
Nice workflow, having an option to repeat a shot a certain number of times would be nice too
tried a lot with 3 shots. all good.
Enabled shot 4 I see the segment does get created - but it's not added to the final video. Only the first 3 are added to the final result.
Near the top of the "Shot 3" section, there is a "Set_final_video" node that is taking an input from the Merge Images node in "Shot 3". Move the "Set_final_video" node to whichever shot you are trying to end on, and take the "Merge Images" "IMAGE" output from that section and connect it to the "Set_final_video" "IMAGE" input node.
It's great for camera angles, but it's not very useful for sex scenes; fellatio doesn't fit well in mouths, and penises get very deformed during vaginal sex.
works great but the face changes on the subject for example on a 3rd shot when a penis is removed from the mouth after a fellatio, but over all good workflow.