I was glad to see someone successfully took on the Talk to the Bot assignment. A few others tried but couldn’t get it to work. As I watched the voice-over with avatar part of the video, I wondered what would happen if you tried to lip-sync a still image. At the tail end of a previous exercise, I noticed that the program animated the Aggressive Technologies logo, so I thought it might work. I uploaded a Van Gogh self portrait to iMovie and stretched it out to 10.6 seconds and shared the file as an MP4. That gave me part of the puzzle. I went to ChatGPT and input “if you could see what I hear” and copied out its response, then pasted that in ElevenLabs to generate a voice. I don’t know what Van Gogh sounded like, so I used a different painter. I took that MP3 and used the Wav2Lip script to combine it with the image. It worked, but it’s a little weird with moving lips on a still image.
I recalled that there was an online tool that would animate still photos in weirdly creepy ways, so I gave that a try. Sometimes it refuses to animate things that aren’t photos, but it worked this time. I downloaded their video and ran it through the lip-syncer. I can’t say that it’s less weird, but at least he blinks. That’s a little more human, right?
I’m kinda glad he doesn’t look at the camera. Based on this experimentation, I may need to revisit my Right Said Freddy Kreuger Mercury video.
