Ok thanks for the reply, I'll give it a try myself ! Huge tx again for you vast.ai template it works perfectly. If you plan to evolve it my 2 feature requests would be :
hello @obsxrver , I wonder if you use LTXV2 (quality / generation time is impressive) and if you plan to adapt your existing loras and also this vast.ai template for LTXV2 ? Thanks
and by the way Nano Banana Pro is impressive at generating synthetic data... I us gemini 3 Pro to generate the prompt for nano banana pro to generate different angles & expressions, all you need is 3 or 4 high quality reference shots...
thanks again @obsxrver your template is working amazingly well, I already have generated 3 T2V loras, I have a question : as my current objective is simply to face swap and I have already my highnoise lora performing well at generating hair color, hair cut and face shape, do you confirm I only need to generate low noise models with better training dataset to improve the face swap fidelity ?
finally I tried on Vast.ai & ~50 min on a H100 to run the high or low with 16 pics, very easy to use, thank you ! I just failed to setup the cloud upload at the end (I ticked the "Upload LoRAs to cloud storage after training" but nothing happen) , I had to scp to retrieve the files
Thanks @obsxrver for this template, I'm fortunate to have a 5090, would it be easy to create a local run version ? I'm currently looking at the code & scripts, I'm not sure how much this is dependent to vast.ai
Thank you
