-
Notifications
You must be signed in to change notification settings - Fork 10.2k
Allow HuMo 14B to work with embedded image for I2V #10034
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: master
Are you sure you want to change the base?
Conversation
This is for the 17B or the 1.7B? |
I only tested for 17B, I can test for 1.7B as well quickly. |
Yeah definitely breaks 1.7B, sloppy on my part, fixing quickly. |
@comfyanonymous Okay I didn't realize that the 1.7B model didn't have the extra channels for i2v, I made some changes that should be much more robust. I tested them and it still works with 1.7B (but without the ability to add a starting image) and the behavior with 14B is still fine too. |
@comfyanonymous @Kosinkadink should I close this one up if it won't be merged? |
Just throwing this out there. I tested it and it works and doesn't affect HuMo basic functionality when not adding I2V embeds. There is a bit of flashing from the first frame, maybe I made a small mistake somewhere so feel free to correct if needed. Image can be added just by chaining the WanImageToVideo node. Can be used for reasonable continuation of HuMo video generations (see 2 video example below with i2v used for second half with the same seed).

ComfyUI_00360_.mp4