Mobius: Text to Seamless Looping Video Generation via Latent Shift

TL;DR

Mobius addresses open-domain looping video generation from text descriptions without annotations by repurposing a pre-trained text-to-video latent diffusion model. The core idea, latent shifting, constructs a latent cycle of length $N$ and denoises with a shifted context so all frames are equally treated, enabling seamless loops of arbitrary length $N$ while maintaining temporal coherence. To mitigate artifacts and extend context, Mobius introduces frame-invariant latent decoding and RoPE interpolation (NTK-aware), enabling stable longer-video generation. Empirical results show improved looping video quality and dynamics compared with interpolation baselines, with a training-free setup and applicability to longer-context video generation.

Abstract

We present Mobius, a novel method to generate seamlessly looping videos from text descriptions directly without any user annotations, thereby creating new visual materials for the multi-media presentation. Our method repurposes the pre-trained video latent diffusion model for generating looping videos from text prompts without any training. During inference, we first construct a latent cycle by connecting the starting and ending noise of the videos. Given that the temporal consistency can be maintained by the context of the video diffusion model, we perform multi-frame latent denoising by gradually shifting the first-frame latent to the end in each step. As a result, the denoising context varies in each step while maintaining consistency throughout the inference process. Moreover, the latent cycle in our method can be of any length. This extends our latent-shifting approach to generate seamless looping videos beyond the scope of the video diffusion model's context. Unlike previous cinemagraphs, the proposed method does not require an image as appearance, which will restrict the motions of the generated results. Instead, our method can produce more dynamic motion and better visual quality. We conduct multiple experiments and comparisons to verify the effectiveness of the proposed method, demonstrating its efficacy in different scenarios. All the code will be made available.

Figures (9)

• Figure 1: Latent Shift for looping video generation. Taking 4 latent toys pre-trained Video Diffusion Models (VDM) as an example, we build a latent cycle and shift the start point in each denoising step in inference for text-guided looping video generation. Notice that, the shifting is conducted in the latent space, we emit the latent encoder and decoder for easy understanding.
• Figure 2: Frame-invariance latent decoding reduces the artifacts caused by the 3D VAE decoding.
• Figure 3: We illustrate this with the example of the toy latent video diffusion model with a context window equal to 4. The utilized RoPE-Interp. enables longer video context without training by interpolation.
• Figure 4: Compare with other methods. We give the first frame, the intermediate frame, and the last frame for comparison. Notice that, both Svd-Interp. and Cog-Interp. are frame-interpolation methods, we manually give the same start frame and end frame as key-frames.
• Figure 5: Ablation study on different latent skip. The shift step in each denoising iteration will also influence the generated content.