When we meet a robot, a virtual assistant, or any character with a physical or digital body, we do not just see its shape; we hear its voice. That sound is not merely a tool for delivering information. It is a powerful signal that shapes our entire impression of the character. A voice can make a machine seem trustworthy, friendly, or frighteningly artificial. It can suggest age, gender, and social status, often triggering deep-seated assumptions about who is speaking and what they are capable of. Yet, despite how much this sound matters, the process of choosing a voice is often haphazard. Designers frequently pick a voice because it is convenient or sounds pleasant in the moment, without a clear plan for why that specific sound fits the character's role or how it might be perceived by different listeners. This lack of planning can lead to unintended consequences, such as reinforcing harmful stereotypes or creating a mismatch between what a character looks like and what it sounds like.

To address this gap, a team of researchers from universities and research centers in Germany and the Netherlands has developed a new framework called the Voice Design Compass. This is not a piece of software that generates voices, but rather a structured guide designed to help creators think through their choices before they ever record a single word. The researchers began by building a prototype of this guide, organizing a series of questions and examples into a modular process. They then invited ten experts from fields like human-computer interaction, sound design, and artificial intelligence to test this prototype. These experts walked through the questions, thinking aloud as they considered how the tool would work in real projects. Their feedback revealed where the guide was helpful, where it was confusing, and what was missing. Using this input, the team refined the prototype into the final Voice Design Compass, a tool intended to help designers create, review, and document a clear plan for a character's voice, ensuring that every sound choice is intentional, ethical, and well-reasoned.

The core idea behind the Voice Design Compass is that a voice is a design material, much like the shape of a robot's body or the color of its casing. Just as a designer would not choose a body shape without considering how it fits the character's job, they should not choose a voice without considering its social meaning. The guide helps designers connect the dots between the character's appearance, its role, and the people who will hear it. For instance, a voice that sounds warm and helpful might be perfect for a companion robot, but the same voice could be inappropriate or even unsettling for a machine meant to deliver emergency warnings. The Compass asks designers to pause and consider these relationships early on. It prompts them to ask: Who is this character? Who will be listening? What kind of relationship should exist between them? And perhaps most importantly, what assumptions might listeners bring to the sound?

The researchers tested their initial version of the guide with ten domain experts. These were not casual users but professionals with years of experience in fields ranging from audio production to ethics and robotics. The experts were asked to imagine they were advising a team on the voice for a new embodied character, such as a museum guide or a laboratory robot. As they navigated the guide, they shared their thoughts on whether the questions made sense, if the examples were useful, and how the tool might fit into their actual work. The feedback was detailed and constructive. The experts agreed that the guide was valuable for forcing designers to reflect on things they might otherwise overlook, such as the ethical implications of a voice or the technical constraints of producing it. However, they also pointed out areas that needed improvement. Some of the questions were too vague or overlapped with each other, causing confusion. The guide felt too heavy in some places, asking for too much information at once, while in other areas, it missed important considerations like the specific physical features of a character or the need for testing the voice with real people.

Based on this feedback, the team made significant changes to create the final Voice Design Compass. They reorganized the questions into four main work areas that follow a logical flow. First, designers define the project and the context, establishing who the character is and where it will be used. Second, they develop the voice concept itself, using four flexible modules to explore the character's function, its physical appearance, the identity the voice should convey, and the specific impressions the voice should create. Third, they review the concept through the lenses of ethics and technical feasibility, checking for potential harms, biases, or practical limitations. Finally, the guide encourages designers to test their ideas, revise them based on what they learn, and keep a written record of their decisions and the reasons behind them. This record, called a Project Record, ensures that the thinking process is not lost, allowing teams to revisit their choices later if the project changes.

One of the most important shifts in the final version is how it handles reflection. In the early prototype, the tool automatically flagged potential conflicts or tensions between different choices, which some experts felt was too rigid or confusing. In the final Compass, this is replaced with open questions that ask designers to identify these tensions themselves. This approach respects the designer's expertise while still prompting them to think critically about contradictions, such as a character that looks friendly but sounds authoritative. The guide also places a stronger emphasis on ethics, specifically asking designers to consider who might be harmed by a voice choice, whether the voice reinforces stereotypes about gender or profession, and how the voice might affect people with different abilities. It does not tell designers what to do, but rather provides a structured space to think through these complex issues.

The researchers illustrated how the Compass works with two fictional examples. In one, a museum team plans a narrator character with a wooden radio body. Using the guide, they decide to give the voice a subtle crackle to match the old-fashioned look of the radio, but they also realize they need to test this effect to ensure it doesn't make the stories hard to understand. They document their decision to keep the effect optional and plan to check with visitors. In the second example, a research team is preparing a small humanoid robot for a study. They use the guide to decide whether the robot should sound childlike or adult, and whether its gender should be clear or ambiguous. They consult the guide's questions about stereotypes and realize that a "gender-ambiguous" voice might not actually avoid gendered associations, so they decide to prioritize clarity and intelligibility for their specific task instead. These examples show that the tool is not just a checklist, but a way to make the invisible process of voice design visible and deliberate.

The study found that the Voice Design Compass is most useful for teams that are early in the design process, or for those who need to justify their choices to others. It is particularly helpful for non-experts who need a structured way to bring voice design knowledge into their projects, and for experts who want to ensure they haven't missed an ethical or technical detail. However, the researchers noted that the tool might be less useful for very simple projects where the voice choice is already fixed, or for situations where there is no time for reflection. The experts also pointed out that the guide is currently based on literature and expert feedback, and it has not yet been tested in a real-world project over a long period. The team acknowledges that the effectiveness of the Compass in actually improving voice designs is still an open question that needs further study.

Ultimately, the Voice Design Compass represents a shift in how we think about the voices of machines. It moves the conversation from "what sounds good?" to "what does this sound mean?" and "who does it affect?" By providing a clear, modular framework for thinking, questioning, and recording, the guide helps designers take responsibility for the sounds they create. It suggests that a well-designed voice is not just a technical feature, but a thoughtful part of a character's identity that respects the people who will hear it. As the field of embodied characters continues to grow, tools like this offer a way to ensure that the voices we build are as intentional and considerate as the bodies we design.