Alpha as an Efficiency Signal: Visibility-Routed RGBA Image-to-Video Generation

2026-08-10

AI Summary

This work addresses the challenges of generating high-quality RGBA videos in a game-like style, which are hindered by data scarcity and the blurriness in translucent regions caused by conventional two-stage pipelines. To this end, the authors introduce GameAlpha-2.4K, the first game-style RGBA video dataset, and propose a single-stage reference-image-guided generation model that jointly produces RGB frames and alpha mattes. Key innovations include a visibility routing mechanism that skips computation for transparent tokens early in inference, an x₀-lock strategy to stabilize denoising trajectories, multi-hypothesis alpha recovery, and a synthesis quality gating module. Built upon a DiT architecture with flow matching scheduling, the method reduces token computation by 35% in two-step inference, achieving a 1.2× speedup while significantly outperforming two-stage baselines in FVD with negligible quality degradation.

Abstract

RGBA videos combine RGB appearance with an alpha channel, enabling animated assets to be applied across arbitrary backgrounds, which are heavily used in gaming industry. However, generating high-quality RGBA animations for games remains challenging for two reasons. First, most existing RGBA video datasets are dominated by photorealistic content, with limited coverage of game assets. Second, the traditional generate-then-matte pipelines estimate alpha only after RGB synthesis, so semi-transparent regions are often blurred by background, resulting in unstable matting outputs. More recently, many methods have begun to model RGB and alpha jointly, but existing approaches are mostly text-conditioned, and still have unresolved issues in efficiency and quality. To address these challenges, we introduce GameAlpha-2.4K, a 2.4K-clip game-style RGBA video dataset built with matte-friendly synthesis, multi-hypothesis alpha recovery, and compositing-based quality gates. Using this dataset, we train a reference-conditioned RGBA video generator that jointly produces RGB frames and alpha mattes in a single pass. To improve efficiency, we propose a visibility router that identifies transparent tokens in an early stage and bypasses their later DiT updates, while x_0-lock guides them along the original flow-matching schedule toward self-predicted endpoints. Our model obtains lower FVD than traditional two-stage pipelines, and the visibility router skips 35% of token evaluations in the final two DiT denoising steps, providing a 1.2x backbone speedup with negligible quality degradation compared to dense inference.

Problem

• RGBA video generation
• alpha matting
• game assets
• efficiency
• transparency

Innovation

• visibility routing
• RGBA video generation
• alpha matting
• DiT acceleration
• reference-conditioned generation