The fact that we don’t already have widely available open source tts executables for Max Headroom and Majel Roddenberry is perhaps the greatest failure of the artificial intelligence movement to date. Frankly I think it’s insane that this isn’t already a thing.
Enter Vall-E which can create TTS models of someone based on just a few samples of their speech.
This link has everything needed to create these models…
Additionally, wav2lip can easily make the clips of max headroom match the tts output. In the future, stability video or other similar ai video generators will likely be able to create more elaborate videos to use for this.