Stable Video Infinity: Infinite-Length Video Generation with Error Recycling
Wuyang Li, Wentao Pan, Po-Chien Luan, Yang Gao, Alexandre Alahi
TL;DR
This work tackles the fundamental drift in autoregressive long-video generation by revealing a training–test hypothesis gap where models trained on error-free histories struggle with self-generated errors at test time. It introduces Stable Video Infinity (SVI) and Error-Recycling Fine-Tuning (ERFT), which re-injects model errors into training to learn error-recycled velocities and stabilize generation via closed-loop feedback. The method employs bidirectional one-step integration and Error Replay Memory to efficiently simulate and learn from error dynamics, enabling infinite-length video synthesis with multimodal conditioning and minimal inference cost through LoRA-based fine-tuning. Across consistent, creative, and conditional benchmarks, SVI achieves state-of-the-art stability, quality, and flexibility for long-form video generation, including audio and skeleton-driven scenarios, with robust scene transitions and identity consistency.
Abstract
We propose Stable Video Infinity (SVI) that is able to generate infinite-length videos with high temporal consistency, plausible scene transitions, and controllable streaming storylines. While existing long-video methods attempt to mitigate accumulated errors via handcrafted anti-drifting (e.g., modified noise scheduler, frame anchoring), they remain limited to single-prompt extrapolation, producing homogeneous scenes with repetitive motions. We identify that the fundamental challenge extends beyond error accumulation to a critical discrepancy between the training assumption (seeing clean data) and the test-time autoregressive reality (conditioning on self-generated, error-prone outputs). To bridge this hypothesis gap, SVI incorporates Error-Recycling Fine-Tuning, a new type of efficient training that recycles the Diffusion Transformer (DiT)'s self-generated errors into supervisory prompts, thereby encouraging DiT to actively identify and correct its own errors. This is achieved by injecting, collecting, and banking errors through closed-loop recycling, autoregressively learning from error-injected feedback. Specifically, we (i) inject historical errors made by DiT to intervene on clean inputs, simulating error-accumulated trajectories in flow matching; (ii) efficiently approximate predictions with one-step bidirectional integration and calculate errors with residuals; (iii) dynamically bank errors into replay memory across discretized timesteps, which are resampled for new input. SVI is able to scale videos from seconds to infinite durations with no additional inference cost, while remaining compatible with diverse conditions (e.g., audio, skeleton, and text streams). We evaluate SVI on three benchmarks, including consistent, creative, and conditional settings, thoroughly verifying its versatility and state-of-the-art role.