Prospective GSoC 2026 contributor interested in Image-to-Video extension for OpenVINO GenAI

[Ganii1729]

Hello OpenVINO team,

I am a prospective GSoC 2026 applicant interested in the “Image-to-Video support for the LTX pipeline” project.

I am currently studying diffusion-based video generation pipelines and exploring how conditioning mechanisms (such as image embeddings or latent injection strategies) can be integrated into existing text-to-video pipelines.

Before drafting my proposal, I would like to clarify:

1. Are there recommended reference implementations or baseline architectures for image-to-video conditioning that align well with OpenVINO optimization goals?
2. Should the primary focus be on architectural integration first, or on inference performance optimization within OpenVINO runtime?
3. Are there specific benchmarking expectations for GenAI WWB integration that applicants should consider early?

I am currently setting up OpenVINO and experimenting with HuggingFace diffusion pipelines to better understand implementation trade-offs.

Thank you for your guidance.