Thanks for the updated model info. Great tutorial. Where are the updated standard workflows located? Or, haven't they been updated yet to include STG etc.?
i'm told there is a training script, but it might need to be done with a big Cloud GPU, so no tutorial until it can train on local, or at least with a reliable service that is not too difficult for artists to use.
I usually use Llava-OneVision, but it does take a long time, so i provided Florence2 as the Vision to caption node here. It's much faster, it is providing "image2prompt" function for the workflow, so you just add an image and let it rip. You can delete the Florence2 nodes and go back to typing the prompt if you wish, no big deal :)
Just came across your channel and workflows this week, thanks for all your work and sharing!!
I
it's my pleasure
Amazing as always. Once we can get Loras, style transfer and Dwpose in LTX we should have a full production pipeline.
That would be cool! I can't wait for the day we see it !!
Fantastic video! Thank you for the workflows! 🙌
Glad you like them!
Thanks for the updated model info. Great tutorial. Where are the updated standard workflows located? Or, haven't they been updated yet to include STG etc.?
links are in description at the top, pick CivitAI or Github, either has the latest workflows as shown in the video, STG is built in now :)
thanks for sharing
My pleasure
It's a great update, thank ^^
Any news for prompt agent for local llm?
Yes - that one is coming soon
Cool
Thank you so much 🙏❤ your workflow works well with my 8gb vram card
Is it possible to train motion luara on ltx to get motion variation if possible do avid on it
i'm told there is a training script, but it might need to be done with a big Cloud GPU, so no tutorial until it can train on local, or at least with a reliable service that is not too difficult for artists to use.
florence? what's the deal with this...why florence?
I usually use Llava-OneVision, but it does take a long time, so i provided Florence2 as the Vision to caption node here. It's much faster, it is providing "image2prompt" function for the workflow, so you just add an image and let it rip. You can delete the Florence2 nodes and go back to typing the prompt if you wish, no big deal :)