I see portrait videos on the cogvideox page
How were these portrait videos accomplished? Was it all post processing or can cogvideox I2V produce portrait now?
https://yzy-thu.github.io/CogVideoX-demo/
This is naturally generated, but currently we do not have control functions such as controlnet.
I get that the controlnet workflow is not available currently. When you say 'naturally generated' do you mean the codebase in this repository support portrait image input (I2V) and can generate output that is portrait ratio (9:16 for example)?
Thank you
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
@yzy-thu will this 16fps model be open-source?
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
@yzy-thu will this 16fps model be open-source?
Stay tuned
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
@yzy-thu will this 16fps model be open-source?
Stay tuned
@yzy-thu actually I already tried 16 fps fine-tuning, but the output is only 3s, I'm wondering how you get 6s video?
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
@yzy-thu will this 16fps model be open-source?
Stay tuned
@yzy-thu actually I already tried 16 fps fine-tuning, but the output is only 3s, I'm wondering how you get 6s video?
we do a patch in T dimension
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
@yzy-thu will this 16fps model be open-source?
Stay tuned
Woah, exciting, you are becoming number one in this space quickly!!! Following with great interest!
This is a model that we continued training based on 5b, capable of generating 16fps 720p resolution videos.
@yzy-thu will this 16fps model be open-source?
Stay tuned
@yzy-thu actually I already tried 16 fps fine-tuning, but the output is only 3s, I'm wondering how you get 6s video?
we do a patch in T dimension
Can you explain it in detail? @yzy-thu