Speedup 4-step LoRa ?
when speedup 4-step lora? is it even planned ? because gen time with even RTX 5090 and any reasonable resolution takes so much time.. thanks
it might match my ltx2.3 gen time in that happens
Maybe we should wait for Lightx2v.
you can do 15 step , is not any diffrent with 20step also use easycache node , for 10sec video only taking 2 minute
How to get easycache node workflow??
It's part of the stock Comfy pack. Put it between your diffusion model loader or GGUF Unet loader and your Sampler.
Info: The MiniMax license is not granting a license to users in EU, US, UK and South Korea (not allowed to run, distribute outputs or modify the model) because of their current lawsuit, and they're doing that to protect themselves. Limits the authors like LightX2V to publish a turbo lora, so it might take a while to get one.
Info: The MiniMax license is not granting a license to users in EU, US, UK and South Korea (not allowed to run, distribute outputs or modify the model) because of their current lawsuit, and they're doing that to protect themselves. Limits the authors like LightX2V to publish a turbo lora, so it might take a while to get one.
oh... that's why Unsloth didn't released a GGUF version yet...
Ostris is already working on a 4-step LoRA for H3. Just have to be patient.