Voiced Play-by-Play and the future of AI in OOTP?

[Speck]

The voice in the sample seems a bit "dry", but I realize it's not meant for something like sports commentary (but still for very helpful for people with vision issues).

I do think AI generated text and audio play-by-play commentary could add a lot of immersion to the game. There could be play-by-play and color analyst teams, with each team having a different feel/personality to their broadcast commentary. Some could be more/less chatty, more/less stats discussion, more/less storytellers, more/less player history discussion, more/less friendly, etc. And as AI voices advance, they could also incorporate those different personalities into the audio broadcast (fixing my "dry" complaint).

From recent Valve posts, it seems like they're fine with AI generated content. They just don't want it going outside of acceptable bounds (sexual, discriminatory, etc.), which this should obviously get nowhere near.

[TheChief]

Originally Posted by novangelus1975

I am loving OOTP 25. So many quality improvements. I do use the tts function when I play and even as stilted as it is, for me it adds to the immersion.

But let me raise my hand along with others in this thread about the value of a tts announcer. Someone raised the issue that Steam might ban the game if it used artificial voices. I think this is unlikely since this is an accessibility issue for many people and not just another bell or whistle.

I've been impressed with the Chris voice-over at Elevenlabs.
I quickly wrote some play-by-play and the voiceover was excellent.
Check it out at https://1drv.ms/u/s!AsFJVvASHzmOlNES...mNDpQ?e=mDZHNl

Until the day arrives when OOTP has its own announcers, is there a workaround that would allow us in the game to choose MS Voices already on our computer? OOTP always plays the Dave voice even if the Mark voice is the default voice on my computer.

You have to prove any Ai generated anything doesn't infringement on any copyright and that the company owns the rights.

So if the game company hired voice actors and built the ai around their voices it is ok.

If it's just some generic Ai voice then it has to proven it was based on voices that are able to be used for sale and not actually real people's voices who didn't give their permission.

Same applies for writing or art. You could load the Ai with the works of John grisham and have the ai create a story but steam would ban it. Because the original rights were not owned.

Now if the company can prove they contracted the author to generate an Ai story it is ok.

Accessibility has nothing to do with it. It is a legal issue over who owns the rights the Ai was built on and can it be proven.

https://store.steampowered.com/news/...63747997849618

[novangelus1975]

Thanks for your reply. I see now your point about the legality of ai generated voices. And I realize that we are talking about different things. With elevenlabs voices, they are created by the studio from voice actors and then digitized. They are not the result of generative AI. One can purchase these elevenlabs voices and legally use them for text-to-speech applications.

[Dementia5]

New to the forum, hello folks!

Just wanted to add that the technology is most certainly there to inject AI generated text for pbp, and also to configure it for TTS that bypasses the Microsoft Sam voice and other Windows speech. Developing such a solution is beyond my coding powers, but I have helped develop and test similar mechanisms for a couple of video games (Skyrim and Crusader Kings), and it works flawlessly on those platforms. I would think with XML being so prevalent within OOTP there would be a means of bridging the two domains.

[PSUColonel]

Originally Posted by tehsuigi

I would happily offer my services as a voiceover guy (DrDynastic would be a great option too). I think OOTPD could manage something like how NHL 2004 does it with base lines and then have the player names as their own miniature files with limited intonations.

I would also...I am a former broadcaster, and did a little play by play at times.

[eauhomme]

Shouldn't be difficult. The main issue would be pronunciation of player names. I remember a game several years ago where they had pronunciations of all the names in their database, then if it didn't recognize a name (say, a newer major leaguer), they would just throw in the position name. "Up now to hit, the third baseman."

[eauhomme]

I could do the voices. I have been told I have a face made for radio...

Seriously, though, I have had my voice used for the phone tree and for training videos for one of my past employers.

[Eugene Church]

A PBP guy with a baritone voice like John Miller would be a dream come true for me.

This would be the final link in a great chain. Natural sounding PBP speech.

[LansdowneSt]

Originally Posted by eauhomme

Shouldn't be difficult. The main issue would be pronunciation of player names. I remember a game several years ago where they had pronunciations of all the names in their database, then if it didn't recognize a name (say, a newer major leaguer), they would just throw in the position name. "Up now to hit, the third baseman."

For historical players, the db would need to include the BBR info on pronunciation, so the Red Sox new reliever is properly pronounced "\ah-ROLL-dis". There are 23,000k MLB'ers that could have this info populated. Just takes some patience to get to them all in the database. Fictional names are tricker but not as jarring if said slightly off. I agree the tech has moved so fast, this is do-able if made a focus point during the development-for-the-next-game cycle for OOTP27-30.

The real voices, be it Harwell, Castiglione or Mel Allen, etc. would be mods. I really look forward to that.

[dodgerblue88]

Understanding AI Pbp would take some time to implement, in the near term how neat would it be to have an AI voice PA announcer? I imagine that should be relatively simple to add in

[OutS|der]

Bad pronunciation of names would add sense of realism.

[Guthrien]

I'm using the 11 voices app, you can plug in any text and it's an audio book and the voices with customization are amazing. It's literally of better quality than many audio books that I own. As someone who worked in an audio engineering business for some time, VO is an art, I cannot believe how talented some of the people we worked with were. That said the current consumer level has just burst through the ceiling over the last couple of years, I do not believe you can tell the difference in most cases and you're able to direct those voices. It will be a cost effective solution in games like OOTP really soon for that reason if they believe it's a worthwhile investment in time.

[Dementia5]

I use ⓍTTS for my text-to-speech services. It is an excellent speech model that lets you clone voices in different languages by use of a 3-second audio clip; you can use your own voice for instance. It comes at my favorite price: free. The trick is injecting the OOTP pbp text into the engine. THis would require some code (like an external running exe), and would probably make use of XML output to accomplish.

[jksander]

Originally Posted by Syd Thrift

Other games do have the Microsoft voices doing PBP. I think maybe you might be able to get OOTP to use them as well, I'm not sure. This is useful from an accessibility standpoint to be sure: people who have eyesight issues should still be able to enjoy the game. If I'm being honest, however, I'm usually speeding past the text anyway once I'm like 40 hours into a game like this. The idea that AI is going to add in more flavor text seems like asking a lot, I think, and the cost of this would be to have the game stay connected to the Internet at all times and to wait several seconds per play while the centralized AI comes up with a way of describing the play. And then of course you're going to need to be able to train the AI so that it "learns" what PBP is good and what isn't, and that's in turn going to require some people with a lot of free time on their hands...

On top of that I kind of have a personal, visceral dislike for AI being used in place of people doing creative work. AI is good at looking up stuff on the Internet and after a few seconds making up something that's convincing. It's not good at all at evaluating what it just came up with: as far as it's "concerned" it did its best and it's up to you to tell it if "best" was good or not. And I'm not at all convinced that it's going to actually improve the repetitive nature of the play-by-play; play-by-play is kind of repetitive in real life, too, we just don't notice it so much because it's sitting there in amidst interesting commentary.

For sure ... I'm legally blind (20/300, so what you can see from 300 feet away I have to be 20 feet away to see it) and I can use OOTP's functions without having to use the voice-over PBP, but I love that it's there ... a lot of games ignore accessibility options that allow blind players to fully immerse themselves.