A super simple workflow that'll have you making i2v ltx2.3 in no time. Easy on/off options and chunking built in. I used a few things from other workflows, but wanted something that fit my needs better. 768x1024 24fps 5secopnd video with interlope and 1.5x upscaling takes on average 160seconds on a RTX-5080, but only 78 seconds without upscaling and interlope.
Description
FAQ
Comments (4)
Is something wrong?
LTXVAudioVAELoader node cannot load LTX23_audio_vae_bf16.safetensors in the path "models\vae"
I must put LTX23_audio_vae_bf16.safetensors into "models\checkpoints", then this node can display the model name.
Yes, that is how that node works for some reason.
Yeah, so simple it doesn't work. RTX upscaler node it impossible to install even manually and even if you deactivate it, the workflow stops working when loading the video VAE.
inside your python_embedded folder; open powershell; then run these two lines in order. ./python.exe -m pip install wheel-stub
./python.exe -m pip install nvidia-vfx_
the Nvidia node will then work. For whatever reason it hasn't been addressed by the node creator and requires manual install.