Paper 2603.04379
Helios: Real Real-Time Long Video Generation Model
- Published
- Mar 2026
- Research lab
- Independent
- Citations
- 33
- GitHub
- 2K stars
01 In brief
Summary
Helios is a 14B autoregressive diffusion model for video generation that achieves 19.5 FPS on a single NVIDIA H100 GPU, enabling minute-scale generation without common anti-drifting heuristics (e.g., self-forcing, error-banks) or standard acceleration techniques (e.g., KV-cache, quantization).
It unifies T2V, I2V, and V2V tasks via a unified input representation and introduces Unified History Injection, Guidance Attention, and Easy Anti-Drifting (Relative RoPE, First-Frame Anchor, Frame-Aware Corrupt) to mitigate drifting.
Deep Compression Flow reduces token and step counts via Multi-Term Memory Patchification and Pyramid Unified Predictor Corrector, while Adversarial Hierarchical Distillation cuts sampling steps from 50 to 3.
Infrastructure optimizations (Flash Normalization, Flash RoPE, memory management) enable single-GPU training without parallelism.
Helios outperforms prior methods on short and long videos, and the authors release HeliosBench, a 240-prompt benchmark for real-time long-video generation.
Code, base model, and distilled model will be released.
Limitations include metric inadequacy, stitching flicker, and 384x640 resolution cap.
02 From the paper
Abstract
We introduce Helios, the first 14B video generation model that runs at 19.5 FPS on a single NVIDIA H100 GPU and supports minute-scale generation while matching the quality of a strong baseline. We make breakthroughs along three key dimensions: (1) robustness to long-video drifting without commonly used anti-drifting heuristics such as self-forcing, error-banks, or keyframe sampling; (2) real-time generation without standard acceleration techniques such as KV-cache, sparse/linear attention, or quantization; and (3) training without parallelism or sharding frameworks, enabling image-diffusion-scale batch sizes while fitting up to four 14B models within 80 GB of GPU memory. Specifically, Helios is a 14B autoregressive diffusion model with a unified input representation that natively supports T2V, I2V, and V2V tasks. To mitigate drifting in long-video generation, we characterize typical failure modes and propose simple yet effective training strategies that explicitly simulate drifting during training, while eliminating repetitive motion at its source. For efficiency, we heavily compress the historical and noisy context and reduce the number of sampling steps, yielding computational costs comparable to -- or lower than -- those of 1.3B video generative models. Moreover, we introduce infrastructure-level optimizations that accelerate both inference and training while reducing memory consumption. Extensive experiments demonstrate that Helios consistently outperforms prior methods on both short- and long-video generation. We plan to release the code, base model, and distilled model to support further development by the community.