workflow reorganization and S2V surpported, with easy switch and instructions
also changed and rearranged some of them.
However, I use Chinese so you may not fully understand it. For some of the node titles, you can try to translate.（“配置压缩元数据” represents the original “CONFIG CompressionArtifacts (I2V)”）
Notice: I have changed the models loaded in the【DualClipLoader】 and changed the load device, as well as the noise seed, you have no need to do that
workflow is included in this picture
(I hope to make Deforum videos, but it seems doesn't actually support such kinda style.)
update
the previous version was especially for i2v (in order to control the size easier, and it's the way I like)
now for t2v, you shall connect the two nodes in group【T2V宽高设定】 to "Empty Latent Video（LTXV）".
back to i2v, reconnect the "junction point"(in group 【I2V】) with "Empty Latent Video（LTXV）"
I2V example
when T2V
support setting seconds as float, total frames will automatically fit for 8n+1(take the nearest number, thus output secends won't comletely match input)
now you can set frame rate freely(not necessarily as 24 or 8n)
you can pre-run the node group to preview total frames
new lora added(if needed)(not tested)
Thank you, it's working fine.
When I want to use only the first frame and disable the last frame, it gives errors on some nodes. How can I do this?
When I want to use only the first frame and disable the last frame, it gives errors on some nodes. How can I do this?
as is told in workflow(in the long guide in the note node), change number of guides in LTXVAddGuideMulti to 1
back to flf2v, remember to change back and reconnect the last frame(frame_idx_2 should be -1, idk why, but it is just set like this)
Thank you, it worked without any issues. The original workflow didn't work, but the one you provided is working. Interesting :)
Thank you, it worked without any issues. The original workflow didn't work, but the one you provided is working. Interesting :)
oh well, that sounds strange, the original workflow should be able to run correctly as supposed to
When I want to use only the first frame and disable the last frame, it gives errors on some nodes. How can I do this?
as is told in workflow(in the long guide in the note node), change number of guides in LTXVAddGuideMulti to 1
back to flf2v, remember to change back and reconnect the last frame(frame_idx_2 should be -1, idk why, but it is just set like this)
An index of -1 means "the last frame" (e.g. total frames minus 1).
When I want to use only the first frame and disable the last frame, it gives errors on some nodes. How can I do this?
as is told in workflow(in the long guide in the note node), change number of guides in LTXVAddGuideMulti to 1
back to flf2v, remember to change back and reconnect the last frame(frame_idx_2 should be -1, idk why, but it is just set like this)An index of -1 means "the last frame" (e.g. total frames minus 1).
thx a lot, I think I understand now, then for more frames to control, shall I use positive or negative numbers? or both are acceptable?
這工作流太棒了，太感激了．．5070ti 生成6秒約1分鐘內
不過如果T2v 的素材有穿網格絲襪，畫面似乎會有點破損...不知可以調整哪邊增加細節度?
謝謝
這工作流太棒了，太感激了．．5070ti 生成6秒約1分鐘內
不過如果T2v 的素材有穿網格絲襪，畫面似乎會有點破損...不知可以調整哪邊增加細節度?
謝謝
其实我自己并没有测试过T2V，也许可以试一下把宽高的倍率调高成16或者32？似乎看到有什么地方说LTX2对于画面大小的设置应当为多少的倍数，但是我并不知道事实情况和效果如何。又或许可以等待新的模型？（或许是模型本身或是lora的问题）
Thank you, it's a very advanced workflow. I tried to randomize the seed section but couldn't manage it. How can I do that?
Thank you, it's a very advanced workflow. I tried to randomize the seed section but couldn't manage it. How can I do that?
open the subgraph and change "fix" to "random", or add a "global random seed" node