#AI generated voice

[mcmiatrevelyan]

This is one of my earlier videos and how I imagine if my MC fell into a coma and Sebastian was by her side at the hospital wing.

[polarisgreenley]

"In the Shadow of Our Ghosts" by @betheckart.

Chapter 13. "The Haven" (English ver.) Audio Excerpt.

Full chapter located here on Wattpad.

Thank you for giving me permission to once again bring this scene to life, and I hope that I did some justice for your beautiful writing of Beth's tale 🩵

*Note: I own absolutely nothing except from the undying love for this tale and the hyperfixation upon HL.*

[seriously-mike]

Ożeżkurwa, stworzyłem potwora po raz kolejny. Otóż jak nietrudno się domyślić, zarówno po polsku jak po angielsku mam pewną manierę pisania i mówienia, i tutaj wchodzi moje streszczanie/parafrazowanie wczesnych etapów wojny w Ukrainie. Sam tekst jest śmieszny, ale wpuściłem go w syntezator mowy ElevenLabs, ustawiłem jako głos (polski!) coś nazwanego "Clyde", z tagiem "War Veteran" i spodziewając się najgorszego, kliknąłem "Generate".

Jak usłyszałem efekt końcowy, to można go stanowczo określić jako #odjebawszy. Oryginalny post poniżej:

Opowieści Wojenne: Przymusowy Ochotnik

Co to są za jaja... Ukraińcy złapali chłopka z jakiejś pipidówy nad Oką (130 km od Moskwy!), poborowego ze Strategicznych Wojsk Rakietowych. Czyli, teoretycznie, typa który miał siedzieć na dupie w silosie nuklearnym i czekać, aż trzeba będzie nacisnąć duży czerwony guzik, a zapierdala w pepegach i z zardzewiałym kałachem gdzieś na Charkowszczyźnie. 

W największym skrócie, przyjechali oficerowie i wylosowali przez wskazanie przymusowych ochotników, jak to w Rosji. Z każdego pułku mieli wziąć czterdziestu chłopa, pułków jest pięć więc mieli wyjechać z dwusetką. Wyjechali z trzydziestoma dwoma, bo reszta poborowych spuściła ich na bambus. 

Oczywiście pan pułkownik i tak się zasapał, że mają za dużo odmawiających służby, więc chłopek ma ruszyć dupę albo wyląduje w anclu. 

Tak to na wojence ładnie, w stu pięćdziesięciu chłopa siedzieli w obozie pod Sułyhiwką, piętnaście kilosów od Izjumia, i któregoś pięknego ranka chłopek obudził się i zobaczył, że cały sprzęt stoi jak stał, ale ze stu pięćdziesięciu chłopa zostało... siedmiu. Trzy godziny później przyszedł trzyosobowy patrol, wielce zdziwiony że ich nikt nie zmienił i gdzie sobie wszyscy poszli. 

Któryś z dziesięciu byszczaków wpadł na pomysł "Chodźmy do Doniecka, tam jest większe zgrupowanie naszych." W pierwszej wiosce po drodze zapytali, którędy na Donieck. Jak się dowiedzieli, że po pierwsze mają do Doniecka 120 kilosów, po drugie i tak już Ukraińcy tam dotarli, a po trzecie są w takiej dupie że najlepiej się poddać, to się przespali i poszli do Izjumia się poddać.

[dungeonmastersconsortium]

THE SECRET ORIGINS OF THE DARKEST DUNGEON WORLD: REVEALED

[otiskeene]

Exploring The AI Text-to-Speech Revolution

Theodore, an introvert, worked for a company that wrote handwritten letters for lonely people. Still coping with his divorce, he installed a new AI operating system, Samantha, whose calm voice and thoughtful conversations made him feel understood. Though hesitant at first, Theodore eventually found comfort in his bond with Samantha. As they spent more time together, he even began to fall in love with her. While Samantha wasn’t real, she became Theodore’s closest confidant.

Although this story is fictional, with the way AI is advancing, our interactions with technology may become quite similar. Let’s dive into how AI text-to-speech (TTS) technology is evolving to feel more human and emotional.

What is AI Text-to-Speech?

AI Text-to-Speech is a technology that transforms written text into spoken words that sound incredibly human. By using deep learning, AI mimics human speech patterns, emotions, and accents. This tech has the potential to change how we communicate with machines, making virtual assistants feel more real.

How Does AI Text-to-Speech Work?

AI TTS learns from huge amounts of data, understanding the tone and rhythm of human voices. It then breaks text down into basic sounds and applies natural-sounding voice qualities to create a realistic narration.

Benefits of AI Text-to-Speech:

Natural Speech: AI-generated voices sound almost human, making listening more enjoyable.

Multi-language Support: AI can speak many languages, making it great for global use.

Emotional Voices: It can capture emotions like joy or empathy, making content more engaging.

Consistency: No more mispronunciations – AI TTS delivers steady, clear audio.

Educational Aid: It helps with learning new languages by offering accurate pronunciations.

Emerging Trends:

Voice Cloning: Soon, anyone can create a digital version of their own voice.

Emotional TTS: AI voices will be able to express emotions, making storytelling more dynamic.

Content Creation: AI will make it easy to generate high-quality audio for blogs, videos, and podcasts.

As AI progresses, TTS technology will continue to become more engaging and human-like, changing how we experience spoken words.

[softwaring]

we’re really out here living in an AI dystopian hellscape and this shit has just begun… the future is grim…

[fedoraspooky]

dusting off the ol' running Mabel meme after all these years because i am Tired

[prokopetz]

Asking the voice actors on my video game project to consent to having AI models of their voices created specifically and exclusively for the purpose of making the dialogue say the player character's name out loud at appropriate junctures no matter what the player chooses.

[seriously-mike]

Ta technologia jest, powiem wam, deczko przerażająca. Jak się ma pieniądze, to można sobie skleić nie tylko Makłowicza, ale również cyfrowo wskrzesić takie legendy jak Tomasz Knapik, Jan Suzin czy Marek Obertyn. Z Knapikiem w ogóle robota jest o tyle prosta, że ElevenLabs potrzebuje pół godziny materiału, a gościu nagrał 20 minut TEDTalku cięgiem, plus jeszcze kilkuminutowe wypowiedzi do różnych programów, więc wystarczy nie mieć oporów natury moralnej. Uszyć Suzina jest trochę trudniej, ale nie jest to niemożliwe - archiwalne klipy na YT są. Najtrudniej złożyć model danych Marka Obertyna, chyba że mamy jakąś gierkę w której podkładał głos, i możemy zripować jego linie dialogowe - tu się idealnie sprawdza Fallout Tactics, bo pliki danych to zwykłe ZIPy i można je pruć szybko i łatwo, zarówno pliki audio jak i transkrypcje.

Nawet bez tego historia o przymusowym ochotniku brzmi przekomicznie, bo pierwsze co mi w niej przychodzi na myśl to chujowa acz całkowicie legalna (i "profesjonalna") polonizacja militarnej strzelanki pokroju Call of Duty (myślcie, Cenega Polska rocznik 2001), ale pierdolenie o żwirze w wykonaniu Tomasza Knapika czy Krystyny Czubówny (na dokładkę do Makłowicza) to jest next level.

Każdy musi to zobaczyć, to nie są żarty, figura ojcowska pierdolenie o żwirze czego chcieć więcej

[darch7995]

𝐇𝐨𝐰 𝐇𝐨𝐠𝐰𝐚𝐫𝐭𝐬 𝐋𝐞𝐠𝐚𝐜𝐲 𝐒𝐡𝐨𝐮𝐥𝐝'𝐯𝐞 𝐄𝐧𝐝𝐞𝐝

𝐒𝐓𝐎𝐑𝐘: ► 𝑊𝑟𝑖𝑡𝑡𝑒𝑛 𝑏𝑦 @darch7995

𝐈𝐌𝐀𝐆𝐄𝐒: ► 𝑃𝑖𝑐𝑡𝑢𝑟𝑒𝑠 𝑜𝑓 𝑆𝑒𝑏𝑎𝑠𝑡𝑖𝑎𝑛: @dvinaamesca

𝐌𝐔𝐒𝐈𝐂: ► 𝑇ℎ𝑒 𝐻𝑜𝑢𝑠𝑒 𝐶𝑢𝑝 - 𝐽 𝑆𝑐𝑜𝑡𝑡 𝑅𝑎𝑘𝑜𝑧𝑦

[mcmiatrevelyan]

I couldn't make one without Sebastian reciting it too.

What would your reaction be?

[polarisgreenley]

An audio excerpt from the English version of Chapter 3 of the beautiful work "In the Shadow of Our Ghosts" by @betheckart located on Wattpad.

Thank you for giving me permission to do this audio for you, and I look forward to seeing Beth's story unfold💗

*Note: I own absolutely nothing except for the undying love for this tale and the hyperfixation upon Hogwarts Legacy.*

[seriously-mike]

Oh yeah, the Typecast one reminds me of Final Fantasy X and the first two Metal Gear Solid games, the male voice sounds familiar but I can't quite place it - maybe Fred Tatasciore playing a military officer of some kind.

I tested another service, ElevenLabs, and it's mindblowing. It's multilingual, so I used an originally American English voice model to read my Polish shitpost and it did so without even a hint of American accent, but with absolutely side-splitting mannerisms of an early 2000s first-person shooter dub. The cheapest paid plan also allows building voice models from just a minute-long sample, so the potential for shenanigans is unimaginable: one guy cloned Robert Makłowicz, another used the only extant recording of Józef Piłsudski's voice, yet another combined Robert Culp's lines from Half-Life 2 with the voice transformer feature instead of simple TTS to have him babble nonsensically.

I'm testing various text-to-speech services, and the first and very promising one is Typecast. It's super clunky and I had to go through their voice library (very limited on their free plan) to find one that could do emotions for Agnes, and it's still far from perfect (it's like a bad JRPG dub from early 2000s!), but the fact that it even allows for reading the text with emotional charge is the promising part.

This content was produced with Typecast, an AI virtual actor service. Characters casted:  Jack (The Liege) Olivia (Lady Agnes) https://typecast.ai

And this was generated in the free video editor for Windows 11, ClipChamp. It's terribly basic and the only thing it has in its defense is putting American and British English voices separately. But as you can hear, they're still awkward and mechanical.

Music: "Hopeless" by Jimena Contreras, Youtube Audio Library

Also, there's a slight difference in phrasing between the two recordings - the one made in ClipChamp comes from the published version of "Swift Justice", the one made in Typecast is a new rewrite.

[verenahx]

For all the Sihtric lovers out there I’ve made an audio with his voice; enjoy girlies 🫶

if you like it tell me if I shall do more :)

after not posting since forever I’ve decided to bless you all.

[ai generated]

@sihtricfedaraaahvicius @sigtryggrswifey @whitedarkmoonflower @foxyanon @finanhasmyheart @emilyhufflepufftlk @gemini-mama

[dextine]

What is going onnnnn

[inkabelledesigns]

It's incredibly frustrating and exhausting to have to question every piece of media that passes your screen as to whether or not it was made with generative ai. I don't like that I can't trust people not to use the literal theft machine for their projects. Especially when it comes to indies.

Like, so there's a new indie horror game called Zoochosis, right? I have no stakes in Zoochosis, a game about a zookeeper that's trying to stop the spread of a mutation with the animals isn't really my speed. But I'd heard of the people who made it before, and I wanted to see if they'd grown. The devs behind it are also responsible for another game, Sparky Marky, which used ai voices, much to my frustration. I tuned into a stream today where someone was playing it, and after hearing a little bit of the dialogue from Doc, I asked if they'd hired actual voice actors. The streamer seemed unconvinced that they had. So I immediately turned it off and decided to go hunting instead, because I needed to know if real people voiced in Zoochosis. I needed to know if this dev team learned anything from their last experience.

Thankfully, there are credits for the voice actors of this game, but they're very far down, which is a little disappointing. The common threads between them is that most of them have a Fiverr page, so I imagine they were hired through there, but I can't confirm that. Frankly I don't love Fiverr given they are also run rampant with generative ai right now, it's not inspiring confidence. But a number of these voice actors have full on websites for their voice work that seem pretty legit to me. In the case of Doc's voice, Tim Stephenson, most of his work is in commercial voice over, not character voice acting, so that explains some of what I was hearing. Commercial VO and character acting for games, shows, etc. are different skills with different needs. Just because someone can do one very well doesn't necessarily mean they're perfectly equipped to do both, especially when I didn't see credits for a voice director here. Without a director, you're recording remotely and going off of your instincts, and while many actors can have great instinct for performing, getting a performance that fits the vision of the project still requires some communication.

I think it's too soon to tell if these actors were actually paid for voice acting, of if these are ai voices synthesized from their likeness. The game is so new that it hasn't shown up on their imdb credits or personal website credits yet, and until it does, I'm gonna be a little skeptical. At the same time, there are many explanations for why the voice acting sounded off. The dev team behind Zoochosis has a lot of names that lead me to believe they may not be native English speakers (and I'm fairly certain I've heard that from others too). Writing and editing dialogue for a game in a language you don't speak is NOT EASY! That's a big ask, and if that's what's happening here, then I'd actually be very impressed by the quality of what I'd heard. It's just...frustrating. I expect this from big corporations and companies, but Zoochosis is an indie game. The folks over at Clapperheads are a small team making this thing real. And I'm troubled by the idea of an indie team, something built on passion and a desire for creative freedom, using generative ai. They're not the only ones to do so. Pastra, a YouTuber who loves indie horror, has started a series of videos that play into this found footage/news broadcast horror story, and for part of it, they used ai generated images of photographs of children to represent the victims of the monster. And they defended this by saying that it was better to use these generated images of children than actual real children. And I'm just sitting here confused by that. Because like, the model that that generative ai is trained on is using stolen images of real children to make those fake photos. So not only does it fall under using a child's image without consent, but you don't even know what children you've stolen the visage from. That is not a good look. The same goes for Indigo Park. AI was used for the opening footage of Issac Indigo introducing the park, and that's part of why his facial movements are so janky. Mason from UniqueGeese at least apologized and said he'd do better next time, but like...the chapter 1 revamp came out and that wasn't changed.

It's small stuff and big stuff, but either way, it's not okay. Until generative ai is trained only on data that is given to its developers consensually, it is not a viable tool for creation. It's not ready to be tested until the ethics part is worked out, and even then, I don't think I'd ever want to touch it. Like I get it, game development is hard, and in the indie scene, it's hard to make it all come together. But also, I expect a project built on passion to at least TRY to be creative in its solutions, and for the people making it to know better than to use stolen content. I expect better. Maybe I'm wrong for expecting better. But I expect better.

I'm just...disappointed. It makes me feel like I can't touch anything in modern media. Like I may as well go back and only play games, watch shows/movies, and explore stories made before 2023. At least then I can trust that it was probably done without the use of the theft machine. That's such a hopeless thing, giving up on modern art and all the potential it comes with. I don't want to give up on art. I don't want to give up on people telling meaningful stories in modern times. Please, humans, keep making art with your own two hands. Please give me something to be hopeful for.