LTX-2.3_-_I2V_T2V_Long_Video_Custom_Audio Workflow question
So i am using this workflow currently to make audio synced anime styled music videos. Mainly of a woman playing a guitar or a piano while sitting. One issue I am having is that when i extend the video from the first clip and so on, the quality continues to get worse and worse, the saturation leaves the animation more and more, and it becomes more and more blurred. Is there a fix to this issue?
Yes i been meaning to address that. For now try the Long Video Single Pass one.. that should have far less degradation over time.
I'll update to older 2 Long Videos to have same logic as the single pass one asap ;-)
Thank you so very much
So i am using this workflow currently to make audio synced anime styled music videos. Mainly of a woman playing a guitar or a piano while sitting. One issue I am having is that when i extend the video from the first clip and so on, the quality continues to get worse and worse, the saturation leaves the animation more and more, and it becomes more and more blurred. Is there a fix to this issue?
I havent tested long video gens yet but the issue with colors I have had using not the optimal samplers. Acoording to the LTX papers res2s is mandatory. U can also check out this overview : https://www.reddit.com/r/StableDiffusion/comments/1ro8qdi/ltx_23_official_workflows_and_pipelines_comparison/#:~:text=0.0%20(Distilled%20weights%20baked%20in).%20Stage%201%20Guidance%2C%20MultiModalGuider%20(nodes%20from%20ComfyUI%2DLTXVideo
Yes, I used the res_2s in beginning due to that recommendation above.
But personally i think it gives a bit of an "overbaked" looks, with a bit "plastic" skin etc.
Euler is smoother, but could be a little less sharp. In the end its minor differences though. Res_2s main "strength" is perhaps that it works with lower steps needed
But res_2s is definitively one of the samplers worth considering (and what LTX also recommends using)