Make Anyone Say Anything (or MASA for short 😉) takes a video of someone talking as input and the text you want your A.I. influencer to say.
Optionally you can also add an audio file if you want to use another voice than the one on the video and a face to do the final faceswap.
What the workflow will do is:
1) Create an audio file using the text you want to be spoken and either the voice cloned from the video or cloned from the optional audio uploaded
2) With the obtained audio, it will apply a quick lipsync to the video using Latentsync
3) If enabled it will use Reactor either to enhance the face features (lips and teeth can become a little bit pixelated when using Latentsync) or to completely FaceSwap the face of the A.I. Influencer.
This is the end of endless real life recording sessions, MASA will never "oops" a line of text making you record all the sequence again. Record your talking video once, use it forever by changing the text!
Of course this workflow needs to be use wisely...You could impersonate anyone but please only use input talking videos you own the rights to 😀 (note like in my demo which is only a proof of concept)