Wetlook World Forum

Current time: Thu 27/03/25 18:55:34 GMT

Translate page FROM -> TO

Translate page TO <- FROM

Message # 101201

Subject: [AI] Video Experiment from my memory

Date: Sun 02/03/25 16:32:40 GMT

Name: Hudson

Email:

Website:

Report Abuse or Problem to Nigel at Minxmovies
If you enjoy this forum, then please make a small donation to help
with running costs: (you can change amount)

So there is a third time my wife got wet! I searched the memory banks. I remember going to a party at her friend's house some years ago and there was a photo collage of all their college friends. One pic was of my wife bobbing for apples at some outdoor fair or whatever, and she was soaking wet in this dark green shirt with the Heineken logo on it, of all things. I tried to recreate a little clip of that using AI!

https://imgur.com/a/abXhdPe

It's still quite a process. I use the custom diffusion model I trained to create an initial picture or frame. That's still the same grind it always is. Then I use a fork of the Hunyuan model training framework (https://huggingface.co/tencent/HunyuanVideo) to create a short clip using that initial frame and a short prompt in English. Honestly, the computational cost is enormous....and creating a video is lengthy, or it's really expensive. I'll try it out one day on a souped-up GPU and see what performance is like.

I thought it was quite fun to see how the video models work. Not bad! Still a long way to go, though (notice it put a ring on my wife's finger towards the end of the short clip, even though I hadn't even met my wife yet in this scene ) and until then, you have to be resourceful and know the limitations of the AI medium. So when you operate within those boundaries, you can get some interesting results.

Report Abuse or Problem to Nigel at Minxmovies
If you enjoy this forum, then please make a small donation to help with running costs: (you can change amount)