Predictive voice casting

FINDING THE PERFECT VOICE

Everybody intuitively knows when a voice perfectly matches the message of a story or commercial. And yet finding the perfect voice is surprisingly difficult. We build intelligent algorithms that translate a briefing into a choice of emotionally driving voices.

 

What makes a good voice?

A voice is the most emotionally evocative sound in the world. A good voice is not necessarily a pretty voice; a good voice is one that transmits the emotional content of a message to the target audience in an authentic, natural way.

 

Describing the nature of a voice

Describing a voice starts with understanding what makes one voice different from another. Through our work on more than 30,000 voice castings, we learned to define voices according to three categories.

 

 

Emotions & Personality

This is by far the most important category, as it describes the emotional effect a voice has on the listener. We use two main description model for voices.

 

Personality

The first, ‘Personality’, describes the perceived personality of a voice. Based on the traditional Big Five personality test, we describe consumer perception of a voice in five dimensions:

  • Openness (inventive/curious vs. consistent/cautious)
  • Conscientiousness (efficient/organised vs. easy-going/careless)
  • Extroversion (outgoing/energetic vs. solitary/reserved)
  • Agreeableness (friendly/compassionate vs. challenging/detached)
  • Neuroticism (sensitive/nervous vs. secure/confident)

 

 

We found that the perceived personality of a voice is not particularly affected by the vocal performance. It can therefore be attributed directly to the voice talent.

 

Emotions

Emotions, on the other hand, are highly dependent on the performance. It’s interesting to note that to evoke the same emotions in two different geographical regions, voices can sound very different. Our emotional model describes voices in eight opposing categories:

 

 

Attributes & Associations

Attributes are the easiest to grasp and the easiest to agree on. They include aspects like gender, age, pitch, language and performance types (commercial, audio book, podcast, e.g.)

Associations are learned attributes perceived by those listening to a voice. In non-English speaking countries in particular, the most common associations are dubbed celebrities. The German voice of Bruce Willis, for example, is so memorable that he has become a celebrity in his own right.

We use a combination of our proprietary closed-loop market research and algorithmic aggregation to collect the necessary data points to assess a voice sample precisely. This technique is at the heart of our voice casting.

 

Predicting the perfect cast

Once we learned how to describe a voice properly, we taught our algorithms to recognise and understand the differences between voices as well as the differences between individual perceptions.

Every search query is processed by algorithms and is matched against the three categories, as well as more than 2,000 relevant attributes such as skills, customer satisfaction, moods and technical certification.

Using artificial intelligence and big data, we are even able to match a voice selection to a script.

You can find examples of where we use this technology via our voice casting services, sprechersprecher.de and wundervoices.com. These services have made us one of the biggest voice casting providers in Europe.