Here’s a fun experiment I worked on: de-aging Sylvester Stallone to see how far I could push the realism. I used ChatGPT for image generation and honestly, it’s hands down the best for consistency. I uploaded 10 images of Stallone into ChatGPT and used prompts to specify the exact camera angles I wanted.
One challenge I ran into was that sometimes the head would look too big, or after a few generations, it didn’t quite look like Stallone anymore. When that happened, I’d either start over or manually tweak the results in Photoshop to get it back on track. Once I had all my frames, I brought them into Luma Dream Machine, using the First Frame and End Frame features to add motion.
One challenge I ran into was that sometimes the head would look too big, or after a few generations, it didn’t quite look like Stallone anymore. When that happened, I’d either start over or manually tweak the results in Photoshop to get it back on track. Once I had all my frames, I brought them into Luma Dream Machine, using the First Frame and End Frame features to add motion.





