Tuesday, January 31, 2023

Ai that voice-over ~ with Tom CliffordVO

While waiting to post the BLOG this month, we were presented with a second unveiling so-to-speak of a voice-over we did this past year.

Apparently, it was bandied about that if the sample rate on the VO was adjusted downward to a "Transformer" classification sound-a-like, it would enhance the engagement level. Check it out by clicking the video box below. 

Reminds us of a Twilight Zone episode featuring Lee Marvin as Steel Kelly, manager & handler to an android prize-fighter... a boxer. 

Things go south and Lee winds up the prize-fighter himself fighting his automaton's battle. After all, before becoming a second rate robot manager, Kelly was a professional fighter in the days when humans were allowed in the ring. Only this time, it's Human versus Machine.

This seems close-to-home while us voice actors are in the crux of an Ai versus real people battle for our livelihoods. Much like Lee's broken-down character 'Battling Maxo' does in the episode entitled "Steel"

So, too, do we go from organic human to robotic speech enhancement story-teller (listen below). And just like Mr. Marvin, we're not quite that sure we're all too happy about that. And if that isn't enough, it seems to be happening again just this side of...The Twilight Zone!

Got a forward of this Free-Think article about the In's & Outs of the Microsoft Ai program/app that can build a robo-clone of your voice after just 3 seconds. Microsoft's 'VALL-E' artificial intelligence is capable of mimicking anybody's voice after hearing just three seconds of speech. However, it's typically only as good as its training data, and Microsoft opted to use Meta’s LibriLight — an audio library containing 60,000 hours of speech from more than 7,000 English speakers — to train VALL-E.

This means the Ai’s training set was “hundreds of times larger” than those used to train existing voice cloning systems, according to the research paper.

When VALL-E is presented with a new voice to clone, it breaks the three second audio clip into bits Microsoft calls “acoustic tokens.” Using those tokens and its training data, it can then predict what the voice would sound like saying other phrases.

We posted in our July 2021, BLOG reports about Anthony Bourdain and Val Kilmer's voices being cloned. It was scratching the surface back then, but to a VO-Guy, this Ai stuff was going to be something taken seriously sooner rather than later. Also, every month we're confronted with voice Ai on our Medium (Twitter) page. That platform immediately transcribes the text into words...and at times, it be freaky close to human. And it's program platform is not all that sophisticated.

After listening along with Medium's Ai dude, a lot of the humanizing may come from punctuation. Comma's, ellipses, dashes could be the key to humanizing. What a fitting comeuppance to one's career of impersonations than to be impersonated by a Hal 9000 in the end. Well, you won't hear me singing D~A~I~S~Y anytime soon.

Perhaps that's just wishful thinking. For in last weeks Business section of the Sunday Times, Ai was getting some additional coverage as it applies to the print forum. It read "Bots can write sonnets and sometimes deceive people, but they're not upending our lives". It could be proffered that they meant to say - yet. Because the article went on to say that Ai is now a major player in the game called Diplomacy. It hearkens to such minds as Alan Turing and brain-trusts like Microsoft,  MIT and Meta. Excuse the pun, but we're talkin' Bots here...and not just to negotiate games, write sonnets and make dinner reservations, but a bot that'll leave people feeling as though they were chatting with a person. This particular bot is called: Chat GPT. It was released in November by Open AI, a San Francisco, California lab. They even gave the Bot a name: Franz Broseph. He's amongst a new wave of chatbots that are rapidly moving machines into new territory.      

In the meantime, why not cozy-up to a comfy resampled version of a relatively normal sounding human being's voice that's been Ai'd into a Transformer sounding character. To play, please click the triangle in the video box below.

FaceBook Business Page

516.488.8840 Office
516.448.9900 Mobile

No comments:

Post a Comment