Is the proposed workflow complete?
In the workflow that is in the repository, only images in the preview window are generated during generation. Or is it just me or am I missing something, but shouldn't a video output node be connected to this workflow as a video combine? Or is this checkpoint created only for generating events from a set of images?
You can use a "VideoCombine" node instead of the preview images to save videos. I should probably update the workflow to have the VideoCombine node, but it is a very simple swap. I'll get to it eventually 😀
I'm working on an easy to use workflow with subgraphs. And one for long running videos aswell. I can post them when they are finished.
I can't upload a workflow (.json file) here. So I have to upload it somewhere else when I come home from work today (german time). That is how the workflow would look (specific for first frame to last frame)
:
Hi guys. Do anyone tried extention? I just used the LatentConcat node which is used in wan s2v_native (not KJ wrapper), but there is that 'video s2v extend' node which I think somehow cares the issue : when I use just LatentConcat (with this Mega_v1 workflow), there are some artefacted frames inserted in between the two latents. None of the latent chunks have that artefacted frames, it only appear after concat....
I will provide a workflow in the next days where you can extend the video basicly endless. It is based on this workflow: https://www.youtube.com/watch?v=YZTvL8C_xz4 but simpler to use and with some errors corrected.
Thanx, now I have it in funcion! Chunks are stitched together seamlessly.
I will provide a workflow in the next days where you can extend the video basicly endless. It is based on this workflow: https://www.youtube.com/watch?v=YZTvL8C_xz4 but simpler to use and with some errors corrected.
Is your long video workflow posted anywhere yet? If you need any help testing it let us know. I am anxiously awaiting such a solution for v10 and/or Mega.
I will provide a workflow in the next days where you can extend the video basicly endless. It is based on this workflow: https://www.youtube.com/watch?v=YZTvL8C_xz4 but simpler to use and with some errors corrected.
Is your long video workflow posted anywhere yet? If you need any help testing it let us know. I am anxiously awaiting such a solution for v10 and/or Mega.
I still working on it to make it as simple as possible to use and on the other hand error free. I think it will be finished today in the evening (german time) or tomorrow.
in the meantime you can try mine, it is baked in the video, just drop the video to your open comfyui. Be aware there are loras in the first loop section and separate ones in the video extension section, so when you un-bypass the extension part, you must bypass the loras you don't want to use ;-)
How to use it ... set the '1 batch' number of frames (the note nearby with times and max frames is note for me, for my system, the values will vary on yours) and set the multiplier : 1 = no extension, 2 = 2x legth of 1 batch
etc... Also - in the extension part - write prompt in one row (do not use ENTER) for one extension block, by ENTER you separate prompt for the next block. If you will be satisfied with the result, right click it > save preview.