I've posted on /robowaifu/ in the past to spitball potential ideas. I've been particularly interested in the speech synthesis. Recently, I was introduced to 15.ai and a game called Strip Poker Night at the Inventory (SPNATI). As most of you probably know, 15.ai is partially the result of the collective efforts of /mlp/ to create accurate speech synthesis of MLP characters. SPNATI is a hentai parody of Telltale's Poker Night at the Inventory. You can check it out at spnati.net.
I got the idea the idea to combine these two things, and webm related is a demonstration of what I think is possible. A fully AI voiced hentai game. You could call it "Pony Strip Poker Night at the Inventory," but I dream of also including other characters, not just the MLP ones.
However, there are a few obvious problems:
There are many lines of dialogue. A character could have well over 1000 lines. This number can reduced, as there are a lot of lines for specific interactions with other characters. I spent about 40 minutes to produce the 32 lines used in the webm. That includes the time for me to type in a line, process it, listen to it, tweak the parameters a bit, and then repeat until I got a decent result. It is not practical for me to process thousands of lines of dialogue across multiple characters. Hypothetically, the process could be automated, but without someone checking the output voice line, the quality would most likely suffer. If 30 or so anons were willing to contribute that much time, then I think one character could be fully voiced in a reasonable time frame. The trouble is that I am unprepared to organize a large scale project like this, and I have no idea if there would be enough interested anons who are willing to contribute.
I'd like to reach out across the webring and figure out how many anons would possibly be interested in working on this. I also would need help organizing the project. I have experience coding, so I should be able to handle the modification of SPNATI to play audio files. But the bulk of the work would require many people contributing voice lines and then organizing those lines into a completed product.
Even if it doesn't take off, I think this is a very intriguing example of how speech synthesis technology could be applied to H-games.