- This is a SVI workflow that enables first-frame-last-frame for continued smooth videos.
- What is special about this workflow is that it uses my own custom nodes to save the latent as a file, and then loads the latent into the workflow so you can freely manage what video clip you want to tinker with.
- These two custom nodes are provided along with the workflow; just drag the bvga_nodes folder into your custom nodes folder.
- It's very convenient to use, as you're able to easily adjust prompts, motion loras, etc, for each stage of the video. You can keep experimenting before you move onto the next segment. And can even start back from previous segments by loading the old files!
Credit to https://civarchive.com/models/2232205?modelVersionId=2820076, whose workflows finally got SVI FFLF working for me, and I learned a lot from.
You will need the following SVI LoRAs:
v1.1 features various fixes and improvements.
Notably:
There are now 2 separate workflows provided:
- 3-k-sampler workflow to get as much quality as possible out of the steps.
- MoE-k-sampler workflow, which automatically optimizes the scheduling.
Updated the latent saving and loading custom nodes:
- v2 nodes better capture the full latent data, to ensure quality is retained.
(Both the v1 and v2 nodes are included in the new folder. You can simply replace the old bvga_nodes folder with the new one)
Fixed:
- SVI nodes are no longer used in the starting video segment
(which in certain situations degraded quality)
- latents are no longer trimmed in segments that don't require it
(which was reducing video length unnecessarily)
Description
FAQ
Comments (13)
You may notice some color shift across the segments (at least I have been). It seems to be a quirk with both SVI and FFLF themselves, especially when interacting with certain models (like the lightning LoRAs).
You can address it to an extent by optimizing your sampler settings, using color-match in post-processing, etc. But I haven't found a complete solution.
Works perfectly for extending video. I swapped out the model nodes for GGUF ones and it still works.
The saving and loading is awkward though.
Yeah, the file management is a bit bulky, but was the best setup I could piece together without losing the flexibility I was going for.
How can I generate the content between the first and last frames while keeping the transition smooth?
With FFLF it depends on how different the end-image is from what your video frames currently are. If the image is fairly close, you should get a very smooth transition. While if there are significant differences you'll get a sudden jump.
I personally like to use a few segments without it enabled (SVI handles this well), and then enable it only to make necessary scene/pose transitions or to loop the video.
You want to increase the motion_latent_count on the extension to smooth out the transition between extensions when using FFLF. You do need to increase the overlap on the extend node by 4 for every motion_latent_count you increase it by (motion_latent_count * 4 + 1). 2-3 is ideal to get a smoothed out transition but it all depends on the scene and the magnitude of the changes the end frame introduces.
Hello, I'm trying to understand your wf before I download it... so, I generate one video with FFLF, then your wf take the last frame from that, i enable "continuous video", and I hit Run again? is that how it works?
Generate the first video with the "starting clip workflow" enabled, either with or without "use end image" enabled.
This will generate your first video to build upon, in a folder named batch# (the number being set by the "index" node).
To make continued clips, disable the "starting clip workflow" and enable the "continued clip workflow", while setting the "file batch selected" node to the batch# file you want to pick up from. You can again choose if you want to use FFLF with any of these, by enabling/disabling the "use end image" node.
The workflow will automatically load in the last image of the previous clip (in the file selected), etc. I've tried to automate as much as possible.
@bvga got it, understood, will try it then, thanks!
@bvga additional question... say i generate 2 additional videos starting from the initial one... will i get 3 separate videos I have to put together in a video editor, or will i get a single 15 second video directly out of this wf?
@artificialotaku Each time use the "continued clip workflow" it will generate a video that combines with the previous video (batch# file) you loaded. So no need to combine them in another workflow.
@bvga awesome, thanks!
Thanks for sharing this workflow. I was struggling before but this helped cement it in my understanding!