V7 is amazing.
This is the best workflow yet. I'm very happy to use this workflow. Everything seem to be very well optimized. Running two turbo dareties loras at once with 0.4 and 0.6 weights is a great tweak. So, I just use them as they've set. No need to switch turbo loras all the time anymore. Works with all the usual samplers and most rarer ones too with clear video quality boost and clear audio without any "metallic hiss" . You've found perfect balance between quality and speed. Also 32 blocks H3 SLA attention Sparse seem to really help with the quality too. I get clearly better distance faces and overall better video quality.
Thanks for the message, I really appreciate it as I take a lot of time to make sure you guys have something that will just work out of the box without hassle. I do have an experimental turbo lora now as well if you ever get bored.
I'd be happy to test your experimental turbo lora. Is it PK Parasyte Turbo? I'll test it and share my test results here or in Civitai.
I'd be happy to test your experimental turbo lora. Is it PK Parasyte Turbo? I'll test it and share my test results here or in Civitai.
Yes exactly
Hello , thanks for the great work , but do you have any workflow for motion continuity .
Thanks for the message, I really appreciate it as I take a lot of time to make sure you guys have something that will just work out of the box without hassle. I do have an experimental turbo lora now as well if you ever get bored.
sup PK, first of all thank you so much for your hard work.
i have a question, i used your V7 and loading the models is not 2 times faster, but the itirations became 2 times slower, i have no idea why, i am using match not max, disabled everything that slows down the run and it's still the same, my previous workflow is 2 times faster iterations wise, 2 times slower loading models wise, you think it's because of loading two loras?
i'm running 3060RTX12gb + 16ram + comfy kitchen
thanks.
Thanks for the message, I really appreciate it as I take a lot of time to make sure you guys have something that will just work out of the box without hassle. I do have an experimental turbo lora now as well if you ever get bored.
sup PK, first of all thank you so much for your hard work.
i have a question, i used your V7 and loading the models is not 2 times faster, but the itirations became 2 times slower, i have no idea why, i am using match not max, disabled everything that slows down the run and it's still the same, my previous workflow is 2 times faster iterations wise, 2 times slower loading models wise, you think it's because of loading two loras?
i'm running 3060RTX12gb + 16ram + comfy kitchen
thanks.
nevermind it was my first run, now the speed is back to normal somehow, cheers my man.
Thanks for the message, I really appreciate it as I take a lot of time to make sure you guys have something that will just work out of the box without hassle. I do have an experimental turbo lora now as well if you ever get bored.
sup PK, first of all thank you so much for your hard work.
i have a question, i used your V7 and loading the models is not 2 times faster, but the itirations became 2 times slower, i have no idea why, i am using match not max, disabled everything that slows down the run and it's still the same, my previous workflow is 2 times faster iterations wise, 2 times slower loading models wise, you think it's because of loading two loras?
i'm running 3060RTX12gb + 16ram + comfy kitchen
thanks.
nevermind it was my first run, now the speed is back to normal somehow, cheers my man.
Whats your venv? Pytorch + cu
Ocassionally if you're near vram limit you'll get a slow gen. If you cancel and start again it goes normal.