Style Morphing, our latest feature takes the art of blending to the next level by allowing us to create entirely new in-between voice styles.
Just as blending colors in painting can create a wide range of hues and tones, blending voice softens the edges between different emotions and creates a more cohesive and visually appealing voice experience, solving the challenge of making AI voices sound authentic.
At Replica one of our goals is to innovate on the forefront of AI voice technology and deliver highly compelling AI voices that accurately capture and portray the widest range of human emotion.We were one of the earliest to market with AI voices for videogames that could shout in the heat of battle or convey the depths of sadness. Many of our voices have been widely adopted by game developers because of Replica’s ability to encapsulate emotion.
However, having individual voice styles that convey extreme depths of human emotion posed an interesting UX challenge. For example, the difference in how ‘sad’ versus ‘panic’ sound, as voice styles are so extreme, the transition from one style to another would sound abrupt and break continuity, thereby sounding more artificial.
Style morphing expands the number of styles for each voice dramatically, by introducing in-between styles for pre-existing key styles.
This means, as a creator, you now have the ability to fine-tune your performance and convey nuanced emotions as your characters can now perform in up to 20 different styles.
This feature is also beneficial for our voice actor network as it elevates Replica voices as the prime choice for game developers and narrative teams wanting the best quality and range of performance, including nuanced control over their performance.
How it works under the hood
The underlying technology acts more like a photoshop colour wheel where you can blend any number of styles together in any combination to create new in-between styles, much like you would blend any number of primary and secondary colours and tints in the world of painting to arrive at a unique shade and tint.
So technically, users could create their own recipe that is a combination of existing styles for each of our voices to get the perfect performance. Because of the massive UX challenge that would pose for our users, we decided initially to create a preset palette of over 1000+ new styles across our 120+ AI voices that users can pick from.
We can’t wait to hear what you’ll create with your favourite characters!