Generation Is Compression: Zero-Shot Video Coding via Stochastic Rectified Flow
📰 ArXiv cs.AI
Generative Video Codec (GVC) uses a pretrained video generative model as a zero-shot codec, eliminating the need for retraining
Action Steps
- Convert deterministic rectified-flow ODE to stochastic rectified flow
- Use pretrained video generative model as the codec itself
- Specify generative decoding trajectory via transmitted bitstream
- Enable zero-shot video coding without retraining
Who Needs to Know This
This benefits AI engineers and researchers working on video compression and generative models, as it provides a novel approach to video coding
Key Insight
💡 Generative models can be used as codecs themselves, eliminating the need for retraining
Share This
💡 Zero-shot video coding via stochastic rectified flow!
Key Takeaways
Generative Video Codec (GVC) uses a pretrained video generative model as a zero-shot codec, eliminating the need for retraining
Full Article
Title: Generation Is Compression: Zero-Shot Video Coding via Stochastic Rectified Flow
Abstract:
arXiv:2603.26571v1 Announce Type: cross Abstract: Existing generative video compression methods use generative models only as post-hoc reconstruction modules atop conventional codecs. We propose \emph{Generative Video Codec} (GVC), a zero-shot framework that turns a pretrained video generative model into the codec itself: the transmitted bitstream directly specifies the generative decoding trajectory, with no retraining required. To enable this, we convert the deterministic rectified-flow ODE of
Abstract:
arXiv:2603.26571v1 Announce Type: cross Abstract: Existing generative video compression methods use generative models only as post-hoc reconstruction modules atop conventional codecs. We propose \emph{Generative Video Codec} (GVC), a zero-shot framework that turns a pretrained video generative model into the codec itself: the transmitted bitstream directly specifies the generative decoding trajectory, with no retraining required. To enable this, we convert the deterministic rectified-flow ODE of
DeepCamp AI