Tencent Unveils AI Video Generator Based on Images

The Chinese technology company Tencent has introduced a new open-source AI model, HunyuanVideo-I2V, designed to transform images into videos.

This tool enables the conversion of a static photograph into a video clip with 720p resolution and up to 129 frames (approximately five seconds) using prompts.

HunyuanVideo-I2V is built on the HunyuanVideo model, unveiled in December 2024, which is the world’s largest open-source video generation model, containing over 13 billion parameters.

The tool allows the use of individual LoRA adapters to create unique visual styles and effects, providing users with greater control.

HunyuanVideo-I2V requires a powerful graphics processor—80 GB of video memory is recommended, with a minimum requirement of 60 GB.

Users have reported that an RTX 4090 24 GB is sufficient if dynamic scenes are not involved.

Some test on HunyuanVideo I2V model for on local RTX 4090 24G
If not dynamic movement, performance is fine. pic.twitter.com/3fBfLMuwZp

— TTPlanet (@ttplanet) March 10, 2025

Following the announcement of OpenAI’s text-to-video generative AI model—Sora—many startups have begun developing and offering similar tools.

Previously, DeepMind, Google’s AI-focused division, announced a new model for creating videos, Veo 2.

In October, Meta introduced Movie Gen, an AI generator for creating new videos, editing existing ones, and adding sound to them.

In July, the AI model Kling from Chinese developer Kuaishou became available worldwide.