I have been experimenting with AI voicing the last few months. It started about 3 months ago when I started experimenting with Eleven Labs AI voicing, and using a free account voiced a part of Rachel Meets Ariane, and posted a video on Reddit here. It was not perfect, but if it is anything like the art and writing AI efforts, it was going to get better.

To finish out my “free” allotment I voiced Strip Crazy Eights and posted the results on Patreon. I have since decided to start voicing all my games, and subscribed to Eleven Labs. Yes I tried out other companies that do the same thing and Eleven Labs keeps improving. When I started, all the voices sounded like narration toned voices for audio books, but the acting keeps improving. Sometime next month I will be releasing Something’s In The Air Redux mostly voiced* (free upgrade for anyone who already got the game).

So I thought as a preview I would release the voiced free version of Strip Crazy Eights available here.

It is not 100% voiced. The intro to the game and the finish to the games are voiced, the actual game play itself is not. I initially tried to voice the game itself but the text got very repetitive, so during the game it reverts to what I call “montage mode” where the music plays and you can see the text, but you cannot hear them speak until somebody wins or loses.

For those that don’t like the voices, going to preferences and lowering the “voice” volume will get rid of them.

Fair warning, this was 3 months ago when the voices sounded a bit flat. The voice models have noticeably improved thanks to training.

BIGGER FAIR WARNING! Strip Crazy Eights is the only game voiced so far.

For the last few months I have been voicing portions of Something’s In The Air Redux from the back and working forward. The thinking is that the best voice work needs to be at the beginning of the game. It is too much work and cost to re-voice the ending sections, and I figured since they are the ending sections, the mistakes in the voices won’t get heard too often. The current version of SITAR on itch.io has the endings voiced including the full “Rachel Meets Ariane” section. Version 1.1 with full voicing (or as much as I plan to do) will be available soon.

My rules in voicing is to focus on the NPC characters, giving them unique voices, or at least not reusing voices where characters share a scene. For example Lydia the bass player, and Lexie the waitress have the same voice model, and I can get away with it because they never share a scene together. Veronica the goth chick and Jessie the store clerk do too.

*As for what gets voiced, it is ONLY NPCs. The player character doesn’t get voiced, mostly out of tradition (it is more fun to head canon your characters unique voice to be your own), and because it would triple the number of lines needed to be voiced, as I would have to voice all those lines as a male and as a female. Also narration lines do not get voiced, you can press the “V” key to enable your computer robot text to speech voice to read these if you want (if your computer is set up for it). “Elemental Realms” dialog is not voiced, and deep dialog trees are not voiced beyond the first choice level.

I have the choice of either 1.) cutting extraneous dialogue, 2.) voicing every meaningless line of extraneous dialog, some of which you may never hear, or the choice I made: 3.) leaving lots of less important dialog unvoiced but having it there to read. There are deep levels of discussions while talking to Rachel at the bar, Rachel at dinner, Suki while smoking, Paula while explaining weather control technobabble, the trivia game and the token game that I didn’t feel like voicing as they are rarely seen or not that useful, but you can read them like world building background info from an RPG.

More on the AI voicing as I do more of it. I’m allotted 20,000 words to voice each month, and so far SITAR has taken 4 months of that allotment, and will take a part of a fifth (July) before I can completely finish. I plan to voice Date Ariane Remastered afterwords, and the other two games after that. SITAR has the most dialog of all the games, but I figure at least a year (3 to 4 months each) to get them all done. All will be free upgrade content, no paid DLC.