Single-step Prediction and Action-conditioned Regeneration at the Edge: one-step cloud prediction with on-device decoding and regeneration when the action changes.
The same input image and seed, with predefined action changes every six seconds that are not provided to the cloud in advance. The hexagon strip shows which model generated each 0.75-s chunk of the SPARE video.
Interactive video world models let users explore generated environments, empowering applications from games to robotics. To achieve good quality, video world models often need to run on cloud GPUs, which can be costly for model deployers. Cloud deployment can also lead to higher response latency. A small on-device model can respond locally, but it loses details and drifts over long rollouts. Motivated by the complementary strengths of cloud and edge computing, we study how to deploy a video world model across heterogeneous edge devices and cloud GPUs. Heterogeneous deployment is challenging because current models typically require substantial cloud computation to maintain video quality, increasing serving cost and potentially delaying responses to user inputs. We present SPARE (Single-step Prediction and Action-conditioned Regeneration at the Edge), a simple collaboration method that combines one-step cloud prediction with on-device decoding and regeneration when the action changes. Compared with the original cloud-only generation, SPARE achieves 2.95× the throughput and reduces cloud GPU energy per second of generated video by a factor of 2.90, while maintaining similar visual quality.
A chunk is a short group of video frames conditioned on the preceding video and the current action, such as a movement direction or no movement. Denoising steps predict a clean video latent from a noisy input. SPARE combines one-step cloud lookahead with local decoding and four-step regeneration when the action changes. Both pretrained models retain their weights.
cloud
device
cloudWhile the current chunk is displayed, the cloud predicts the next chunk assuming the current action persists. It sends the x0 prediction from its first denoising step. The device’s VAE decoder converts this clean-latent estimate into frames, so the cloud does not run the decoder.
If the action changes, the device discards the prediction made under the previous action. Its small model regenerates the chunk with all four denoising steps under the new action, without waiting for a cloud response. A device that cannot run the small model at the playback rate keeps only the decoder and leaves regeneration to the cloud (SPARE without the device generator).
Every displayed chunk enters both models’ generation histories, so the cloud continues from the regenerated chunk. More frequent action changes reduce the number of cloud predictions that can be reused. No cache tensor is transmitted between the models.
On 300 WorldScore scenes with actions changing every six seconds and no access to future actions, SPARE scores 7.07 out of 10 under a Qwen3-VL-32B evaluator, against 7.09 for four-step cloud-only generation. Throughput rises from 7.3 to 21.4 FPS (2.95×), and cloud GPU energy per second of video falls from 1,172 to 404 J (a factor of 2.90). These ratios use unrounded measured means. The total GPU energy of SPARE, 655 J, is higher than the 567 J of one-step cloud generation. Without the device generator, throughput is nearly unchanged (21.3 FPS) and VLM quality is 7.03, so local generation does not explain the throughput gain over four-step cloud generation. All results are for LingBot-World on workstation GPUs with injected network delay and 16-FPS playback.
Longer and repeated runs. On 20 scenes × 10 seeds, SPARE and four-step cloud-only generation score 7.27 and 7.40 in VLM quality. For 120-second rollouts (12 scenes × 3 seeds) they score 6.61 and 7.00. Both gaps favor the four-step cloud model.
Every method receives the same input image, actions, and seed. Great Wall and Stylized forest are the two rollouts of the qualitative figure in the paper, chosen by the gap in the VLM drift score: the device model, Hybrid SD, and SDVG lose the wall and the stylized trees, whereas the cloud model and SPARE keep both scenes to the end. The visible wall and trees illustrate appearance over time; they do not by themselves establish action accuracy. EC-Diff, which is not in that figure, keeps the wall but not the stylized trees. Lake and Alley in the rain are two further scenes of the 20-scene set.
Sixteen scenes of the 20-scene evaluation set, seed 1234, without further selection (the four block-world game openings are left out). Hover over a scene for a preview of SPARE, click it to compare.
Local regeneration makes the first decoded frame available 0.20–0.41 s earlier than without the device generator, but the player’s buffer hides most of this gain in display latency. On WorldScore-300, a 2-s network outage reduces stalls from 1.33 to 0.29 s per clip, at 723 instead of 668 J total GPU energy per video second. These are injected-delay and outage tests on one host with a 16-FPS player, not a wide-area deployment, and outage video quality has not been scored.
Older appendix test on the 20-scene set, not WorldScore-300. With cloud messages withheld for 20 s, playback stalls for 0.29 s per clip with the device generator, compared with 19.33 s without it. This tests playback continuity only, not video quality during the outage.
We vary cloud sampling depth, device computation, and generation history. On LingBot-World, the one-step 14B model scores 7.03 against 7.09 for four steps and 6.49 for the four-step 1.3B model. On the 20-scene analysis set, the small model’s regenerated chunks are sharper with cloud-generated history than when it generates the rollout alone. We use one cloud step for LingBot-World and the small model for regeneration after action changes.
We apply SPARE to Self-Forcing 1.3B on the device and Krea Realtime 14B in the cloud, both based on Wan2.1, with text prompts as actions that change four times per rollout and no advance access. The VLM quality score is 7.95 against 7.60 for one-step cloud (95% CI of the difference [0.26, 0.44]), although the reference knows future prompts; final-segment MUSIQ is 66.1 against 63.6, while whole-clip imaging quality stays below four-step cloud. Without the device generator, throughput falls from 9.8 to 8.8 FPS and cloud energy rises from 844 to 1,025 J. Four device steps take 0.48 s versus 0.94 s for one cloud step. Throughput remains below the playback rate, and this dynamic-prompt test has no network or playback clock.