Lights, Camera, Carbon: Architectural Scaling Laws for Video Generation Energy Consumption

2026-07-05

AI Summary

This work addresses the lack of interpretable and standardized methods for evaluating the energy consumption of text-to-video generation models under unknown architectures and parameter counts. The authors propose a weight- and implementation-agnostic bidirectional estimation framework grounded in first principles and observable generation parameters—such as output resolution and duration—to establish, for the first time, a theoretical scaling law linking model energy consumption to architectural complexity. By decomposing the quadratic and linear energy components inherent in diffusion model inference and integrating empirical GPU measurements with observed inference times, the framework retrodicts architectural efficiency and enables a unified sustainability benchmark across diverse models. Validated on six open-source models (8.3B–27B parameters) across three GPU types, the approach achieves an average absolute percentage error below 3% in energy prediction, accurately capturing architectural differences.

Abstract

We present a bidirectional framework for estimating the energy consumption of text-to-video (T2V) and text-to-video-audio (T2VA) models from architectural first principles and observable generation parameters such as resolution and duration, requiring no access to weights, model size, or implementation details. Forward, it predicts energy from generation parameters and architectural principles; backward, it recovers architectural scaling behavior from observed inference times, with accuracy serving as a criterion for architectural validity. Building on the established compute-bound nature of video diffusion models, we demonstrate that each model's energy profile obeys theoretically derived scaling laws, decomposing into quadratic and linear terms whose coefficients directly reflect the underlying architectural complexity. Validated across six open-source models spanning 8.3B-27B parameters and three GPU configurations, this decomposition achieves below 3% MAPE across all architectures. This approach offers a standardized, empirically and theoretically grounded framework for sustainability benchmarking across T2V models and architectures.

Problem

• energy consumption
• text-to-video generation
• architectural scaling laws
• sustainability benchmarking
• video diffusion models

Innovation

• energy consumption
• scaling laws
• text-to-video generation
• architectural complexity
• sustainability benchmarking