Artificial Intelligence 10 min read

High-Fidelity Image-to-Video Generation for E-commerce with AtomoVideo and Noise Rectification

Alibaba’s AI team introduced AtomoVideo, a diffusion‑based image‑to‑video generator enhanced by a training‑free Noise Rectification module that adds and corrects controlled noise to eliminate first‑frame errors, enabling merchants to automatically create high‑fidelity 4‑second 720p product videos with strong temporal consistency for e‑commerce advertising.

Alimama Tech
Alimama Tech
Alimama Tech
High-Fidelity Image-to-Video Generation for E-commerce with AtomoVideo and Noise Rectification

In the e‑commerce domain, video content is becoming a key medium for advertising, but traditional production is costly and labor‑intensive.

To address this, Alibaba’s AI team has developed AtomoVideo, a diffusion‑based image‑to‑video generation system, and a training‑free Noise Rectification module that improves fidelity without extra model training.

Noise Rectification adds controlled noise to the input image, then corrects the predicted noise during DDIM denoising, eliminating the first‑frame error and preserving product details.

AtomoVideo upgrades the base T2V model to an I2V model with multi‑granularity image injection, high‑quality dataset construction, and progressive motion‑intensity training, achieving 4‑second, 720p videos with high temporal consistency.

The system has been deployed in Alibaba’s advertising platforms, enabling merchants to generate dynamic product videos automatically, with examples shown in the article.

Future work includes extending video length, improving stability, and exploring larger‑scale AIGC models such as Sora.

e-commerceAIdiffusion modelvideo generationAIGCimage-to-video
Alimama Tech
Written by

Alimama Tech

Official Alimama tech channel, showcasing all of Alimama's technical innovations.

0 followers
Reader feedback

How this landed with the community

login Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.