Skip to content

Initial implementation of Frames-to-video#1262

Open
gravis778 wants to merge 6 commits intodeepbeepmeep:mainfrom
gravis778:main
Open

Initial implementation of Frames-to-video#1262
gravis778 wants to merge 6 commits intodeepbeepmeep:mainfrom
gravis778:main

Conversation

@gravis778
Copy link

This is under the wan2.2 models. Right now it only accepts start frame and end frame - it doesn't have intermediate frame support yet. I am unsure if this is working correctly - it DOES produce a video that seems to morph the two images, but this doesn't look quite like the examples on the frames-to-video page. I was unable to get the code to run without wan2gp as I needed the memory management to get it to run. This was an initial attempt to get code to work inside of wan2gp, so I am sure there is something I missed.

gravis778 and others added 5 commits October 12, 2025 23:35
…CE workflows

- Added `inject_from_start` and `injection_denoising_step` flags to control early latent injection
- Inserted fallback definition for `source_latents` to prevent UnboundLocalError
- Initialized `randn` tensor before injection logic
- Changes are scoped to VACE and i2v model types; safe for non-injection workflows
Added injection_denoising_step, after denoising_strength
spinal-cord added a commit to spinal-cord/Wan2GP that referenced this pull request Dec 29, 2025
spinal-cord added a commit to spinal-cord/Wan2GP that referenced this pull request Dec 29, 2025
@deepbeepmeep
Copy link
Owner

Thanks but it is not clear based your comments, if this works or not

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants