Tencent just released its open-source HunYuan Image-to-Video AI model, and it got many people excited, so I tried it out on my computer yesterday. These are the few videos that I found acceptable enough to share here.
A square video with 512p resolution takes a few minutes to generate on my desktop PC with an Nvidia RTX 4090 GPU. If you bump up the resolution to 1080p, it takes about 20 minutes. And even after all that waiting, you get an AI video that’s…okay lah. The original still images were generated in Midjourney and then uploaded to Hun Yuan I2V.
As I told my Gen AI workshop participants at The Straits Times Masterclass and at Singapore Business Federation this week, AI videos are still in their infancy and marketers should not spend too much time on them. The tech will get better in coming months but for now, they require too much time, compute and eyeball-rolling.