Scott Lighthiser showed a novel way of using various AIs for human face transformations in videos.
With the continuous development of various text-to-image and image-to-image AIs, it is hard to surprise anyone with the picture-altering capabilities of different apps. Altering videos, on the other hand, is another thing entirely, with many of the most popular AIs being unable to produce a consistent result without any artifacts just yet.
Thanks to filmmaker and musician Scott Lighthiser, however, one possible way of video-altering has been found, and it involves three separate free-to-use apps – Stable Diffusion, EbSynth, and Koe Recast.
Using Stable Diffusion, Stability AI's recently-released program with both text-to-image and image-to-image capabilities, the author turned his face into Greek-like marble sculptures and even a blood-soaked zombie. The AI-generated frames were then applied to the motion information of the pre-recorded videos with the help of EbSynth, a program capable of transforming a video by painting over a single frame. According to the creator, the key was to keep movement to a minimum so as not to add new pixel information to EbSynth. To alter his voice, Scott utilized Koe Recast, an AI-powered app that allows you to transform your voice in real time, letting you alter the core characteristics of your voice to sound however you'd like. Combining all of those programs, the author managed to produce an altered video that depicts statues and a zombie reciting the "whatever the weather" tongue twister.
And here's a similar experiment by Scott, also made with SD and EbSynth:
You can learn more about Scott's other projects by clicking this link. Also, don't forget to join our Reddit page and our Telegram channel, follow us on Instagram and Twitter, where we share breakdowns, the latest news, awesome artworks, and more.