From still image to talking robot avatar in one afternoon

In a single afternoon: from a still image to a talking robot avatar for our AI workshop, the whole process including cost and surprises.

AI avatars keep getting more relevant, but how do you actually build one? Here is our process for Robi, who hosts our workshop:

1. Image generation:↳ Create an expressive robot image in Midjourney. ($10 per month)

2. Script:↳ Wrote the text for Robi's workshop intro and outro with Claude. (€20 per month)

3. Voice:↳ Picked a fitting voice in ElevenLabs and adjusted it. Interestingly, an Indian voice worked best for our German text! (€0)

4. Movement instructions:↳ Uploaded the photo of the robot to Claude and had it describe fitting movements

5. Animation:↳ Combined the image, the movement instructions and the audio file in Hedra, and Robi came to life.

Challenges? Questions in the script did not always sound like questions, which called for creative workarounds. We often had to help along with prompted pauses. Every version carried different emotion.

And at roughly €5-6 per 70-second video, Hedra is not cheap, but so far it is the only tool we know of that animates robot avatars from a single image convincingly.

Many thanks to Esther, who brought Robi to life with her creativity and let us watch her process. That was the basis for our next steps.

The whole process was finished surprisingly fast and was enormous fun! Robi went down very well at the premiere of our workshop. He welcomed the participants and said goodbye to them.

I am curious how workshop participants react to him once we give him bigger roles!

What experience have you had with AI avatars, and what would you use them for?