This is my workflow for wan2.1 works pretty good
Description
this is the image to video version
FAQ
Comments (12)
Thank you! The very first img2vid workflow that I actually got running on my machine. Little bit slow. 16Vram, 720px, 160 frames: 6 hours.
I have replaced the captioning by a normal positive prompt field. Works better for me.
Would be great if you could bring in some loras
I need to learn how to add them. My workload is crazy. Did you try using the 1.3billion parameter model for this. I tried with my 4090 and it took 90 seconds
OK im running it through now to see if it works.
You can just buy 4090 with the electricity bill youre going to have
@NoneofYourBusiness2 already have the 4090 lol
@NoneofYourBusiness2 Where I live, a 4090 costs as much as a used middle class car. But at least I don't have to worry about energy. Running on solar cells.
@2thecurve thanks for the hint. Will try that next.
@perfectboobcollector I need to upload my text to image workflow I have
@2thecurve I have tried this one. wan2.1_t2v_1.3B_fp16.safetensors
Speeded up from 6 hours to 50 minutes but the result had nothing to do with the original picture. Could you give me the direct link for the correct model? Perhaps I took the wrong turn.
@perfectboobcollector yeah that's for a text to video not an image to video. When I get home I get the text to video workflow, uploaded it. I looked for a 1.3b for I2V and apparently it doesn't exist yet. Or I can't find it anyways.
@2thecurve found this one but it also takes about 5 hours to complete.
https://civitai.com/models/1295569/wan-ai-wan21-video-model-safetensors-andkijai-workflow-included?modelVersionId=1463630
I just reread this. It takes you 6 hours for one 8 second+ video? Have you played with forge it isn't like this but it's more doable that 6 hours. Maybe 30 minutes