This post originally appeared on Medium, Jun 20, 2019.
We’ve developed an AI that can replicate the human voice with a few minutes of speech. The following video was made featuring Replica Voices:
How did we do that?
We trained our AI on audio from YouTube interviews of those artists and celebrities, and created the voices you heard in this video with our custom text-to-speech tools.
This video demonstrates how powerful AI voice technology is and also how easy it is for us to produce any number of Replica Voices with limited source audio.
The more audio data we train on, the higher quality and smooth the voices will sound.
Our goal is to empower creative people with 1000’s of AI voice actors on demand.
This technology has huge potential as a creative tool. The next generation of games, movies, and ads will benefit from the scale and versatility of AI generated voices like these.
On the other hand, it’s scary to think of the dangerous implications AI voice technology enables. It’s an obvious target for hackers, scammers, and identity theft.
So we want to talk about what we are building, why we’re doing it, and how we can ensure that people’s rights and privacy are protected as we enter this era of synthetic media.
We want to bring the joy of creativity to everyone.
Before CGI and visual FX became mainstream, film and movie studios got creative with cameras, lighting and mirrors to produce special effects.
Creativity is evenly distributed; talent and opportunity are not.
With the advent of digital tools used by studios, we’ve seen an explosion in the number of animations, games, and movies.
Synthetic AI tools are an evolution of that digital trend. AI further expands the tool set available to creative studios around the world, whether it’s a pair of teenagers in a basement working on an indie game, or massive studios with multi-million dollar budgets working on a major block buster movie.
Replica is launching Studios — a product that enables anyone to replicate their own voice, or voices of their game, movie, or animation characters and create with them.
Studios is aimed at creatives. Users can type up scripts and dialog, and instantly play it back in the voices of their characters, with easy-to-use style filters that control the mood or emotion of the voice.
You can express any emotion or feeling in your Replica Voice and fine tune it until it’s just right.
The creative possibilities offered by this to artists, game developers, animators, and advertisers are enormous.
Stories and scenes can be developed, and character dialogue can be produced rapidly and inexpensively.
Whether you’re an indie creator or AAA studio, you have the ability to rapidly experiment with instant re-takes, and sculpt the emotion, tone and intensity of the voice. It’s like Photoshop for your voice.
Video game developers will use Replica Voices to experiment with script ideas, and create immersive games where dialog evolves with your playing style, making each player’s experience unique.
Advertisers can create personalised radio and podcast ads, or promotional material that is relevant to the individual listener — in the voice of the artist, advertiser, streamer or podcast host. You can scale voice ads, just like any software, and create new revenue opportunities.
The most interesting opportunity offered by this technology is the ability for individuals and artists to licence their own voices at scale — creating new revenue streams for themselves.
And they can do this, without spending time in a recording studio, because their voices can now scale like software.
Imagine buying an audiobook and paying more to have it narrated in the voice of Stephen Fry. Or the calm, soothing tones of Morgan Freeman’s voice for your favourite meditation app. Or maybe you’ve bought tickets to attend a music event and every fan gets a personal message in the artist’s voice.
The possibilities are immense, because Replica Voices scale like software.
Today we’re working with several artists, game, and animation companies to help them scale their voices. We invite you to sign up for early access, so you can try it out for yourself!
Replica Voice Marketplace
Replica is building a marketplace of the world’s voices where artists, individuals, rights holders and publishers can licence their Replica Voices at scale, while retaining full control over where and how their voices are used.
Launching our consumer product — Studios, is step 1 on the journey to a voice marketplace.
We want to give everyone the ability to create at scale. You can use your own voice, or voices that you own the license to (eg: voices for your game or movie characters).
Artists can reach fans more effectively by engaging on a personal level. Users of our software can collaborate and share Replica Voices with people they trust to work on creative projects together.
The ability to scale voice inexpensively will expand the market for creativity with voice, and produce new revenue streams in media and publishing.
Not all voices are created equal. Some voices are more valuable than others.
This is a great opportunity for artists and voice actors to licence their voice at scale. They could earn 10x or even 100x because they can essentially be in 100's of places at once.
Privacy and Security with Synthetic Media
“Your voice, is your identity. And it can be secured.”
While we take privacy and security very seriously, we do need to face some hard facts. A lot of the AI research for ‘synthetic voices’, or ‘synthetic faces’, is open to the public. Today it is still quite complex to understand and implement.
This will not always remain the case as young, talented engineers around the world develop skills in AI and Deep Learning, and the pace of research continues to accelerate. When it comes to your own voice, your identity is exposed to a whole new level of risk.
So how do we secure our own voices?
Remember when transferring money across the world used to be a scary concept? Yet today we take it for granted.
The reason we trust brands and companies like PayPal and Stripe with our money is because they attract the best engineering talent to work on the side of good. In addition to processing transactions efficiently at scale, they invent new methods to enforce privacy and security, and keep ahead of malicious hackers.
Replica provides a secure method to license and monetise your voice at scale.
Our plan to protect voice in the new world of synthetic media:
- You own all the rights to, and control your own Replica Voice(s), and you decide whether or not you want to licence it out — and if you do, whom you licence it to and what they can and cannot use it to say. Think of how AirBnb makes it easy for anyone to rent their room out to strangers.
- We watermark every Replica Voice created by our software so it can be traced back to it’s source. We can prove whether or not a voice file was produced by us, and we’ll know who it belongs to.
- We will work with publishers and platforms like Youtube, Spotify, Audible, Twitch, Activision, etc. to help them enforce digital rights — think of the ‘organic’ sticker on a piece of fruit, we help distinguish between ‘licensed Replica’ or ‘illegal fakes’. We will partner with distributors to provide end-to-end security for users of Replica software.
Our team is committed to tackling the privacy and security issues we all undoubtedly will face in this new era of synthetic media.
We have an important job in educating the public and in taking active steps to ensure we have the world’s best engineers working on these hard challenges.
If you share our sense of responsibility and want to tackle these challenges with us — introduce yourself in a quick email to email@example.com.
Who is Replica?
Replica is a Techstars Company. We recently attended the Techstars Music accelerator program held in Los Angeles, and have returned to our HQ in Brisbane, Australia. We’re glad to have the powerful network of Techstars Music with its numerous member companies and mentors, many of whom are among the biggest names in music and entertainment, globally.
We are really glad to be backed by Mawson Ventures — Australia’s leading AI Lab who have been immensely helpful with financial, executive and product support.
Our team members are passionate gamers and movie buffs. We want bring the joy of creativity to more people. We’re deep learning and cyber security engineers with years of experience shipping amazing products.
Replica is building a marketplace of the world’s voices. We want to help artists and individuals licence their voice securely, and create with voice at scale.
We look forward to comments and feedback on our product. You can reach us directly at firstname.lastname@example.org
— Team Replica