Influencers should be more scared of AI than software engineers

The more AI advances, the more we hear about different jobs going away and will soon be automated.

The advancement of coding LLMs such Google’s Codey or Meta’s Code Llama made it appear that it was only a matter of time before the jobs of software developers and engineers are no more.

If you look at demos, there isn’t really a use cases that has proven AI can replace a specific job, however recently, a research group out of Institute for Intelligent Computing/Alibaba Group showed off the ability to animate any standard image titled Consistent and Controllable Image-to-Video Synthesis for Character Animation.

This new innovation could make influencers on TikTok and other social media platforms obsolete and allow brands to create their own personas and generate content all through AI.

While there is no code posted yet, the demo shows some promising ability to take a still image and animate that image info a full video using an attached wireframe.

Outlined below is how a reference image which can be a simply self of a person or an AI generated person. The pose sequence is a wireframe model that can turn and dance representing how the image should be animated, for example by rotating left and right as a fashion blogger would pose or model a specific outfit.

Once generated, the output would produce a short clip of the image taking form of the wireframe pose and producing a very good video of the subject.

While there are some flashes of mysterious hands that have plagued AI generated content well as some screen flicker and quality issues, this has come a long way in generating new content from still images.

Read the paper here or check out the video on YouTube to see the model in action and comparison of it side by side to similar models like DreamPose, BDMM, and DisCo.

Thank you for reading this version of the Techies Blog!