Cartoon, human-like or realistic? - New research on avatar styles

Which avatar style is right for your application? Which virtual appearance is accepted by your target group?

Short on time? Scroll to the bottom of the article to read the short version for those in a hurry.

Avatar styles

At Charamel, we believe in innovation and progress through research and have dedicated an internal research effort to answering these questions. If your first intuitive thought is that realism is most preferred, you may be surprised by the results.

First, we must consider the three foundational elements of target audience acceptance, avatar appearance, and use case applications. Target audience acceptance criteria include individual socio-demographic characteristics such as age, gender or expertise, but also subjective experience and perception criteria regarding the use case application and the avatar. The focus is on aspects such as credibility, trust, usefulness, or ease of use (usability). In addition, the type of task and situation are decisive, which form a contextual framework for users’ assessments. Habits and social conventions may already prescribe guidelines for an acceptable appearance, as exemplified by the "mandatory suit" worn in the financial sector. A look should also be taken at the requirements for the use case applications. Here, a distinction must be made as to whether the system is a simple one that leads to the goal in an efficient and effective manner without much effort, or whether, in addition to pure functionality, more complex emotional and relational needs must also be met that require more than superficial interaction.

For the time being, the habit of interacting with other humans suggests acceptance and sympathy regarding realistic rather than artificial avatars. But in terms of the Uncanny Valley effect, more abstract representations are advised instead of realistic visualizations. This phenomenon from researcher Masahiro Mori from the 1970s states that robots with human-like appearances seem scary to people and they can feel deceived when they learn that the machine is not a real human. If the machine approaches a lifelike appearance without fully achieving it, empathy turns to disgust. Only full humanity in terms of appearance and behavior would imply acceptance. The complexity of situations with interactivity is not considered, but behaves independently of appearance. Thus, coherent interaction by means of human communication can circumvent the Uncanny Valley effect. That is, communication causes the target audience to evaluate the avatar based on more than appearance. Therefore, the appearance of a virtual assistant can be made realistic if users prefer, but the avatar should clearly communicate that it is not a human but only a machine with limited functionality to avoid false expectations.

In our research, 25 women and men between the ages of 21 and 75 with a high school or college degree and with and without a media/IT background were interviewed about the three avatar types - cartoon, human-like, realistic - in two sample scenarios. On the one hand it was about functional information at the train station terminal (simple system), on the other hand about knowledge transfer in a training environment (complex system). Taking into account the influencing criteria described above, it should be emphasized that the respondents expressed very different opinions in some cases.

The cartoon style was praised for being lively, easy-going, cute and fun, so that questions could be asked without hesitation. The intentional artificiality made it clear that the intention here was not to replace a human being. However, most respondents attributed childishness and a lack of seriousness to this style. For optimal acceptance, the comic style, which is perceived as lively and casual, should be used in childlike, playful areas, suitable for children, light topics and simple information at the terminal.

Human likeness can appear better than realism, also with reference to the Uncanny Valley effect, because in this way a superior and identity-preserving position of the human is maintained. While the comic style is assigned fewer human competencies due to its artificiality and the realistic style is assumed to have extensive human competencies, the human resemblance is positioned exactly in between, so that simple to complex tasks are conceivable here. With the help of suitable clothing, a good middle ground can be found for human-like avatars in order to appear professional, objective and serious without appearing too strict and unapproachable. A majority of respondents felt that human-like characters were more trustworthy, credible, competent, and reliable than cartoon avatars.

Realism divides opinions. While cartoon characters are not perceived as real social actors, photorealistic avatars can be mistaken for real communication partners, so there is a risk of confusion with true interactive situations. Through realism, typically human competencies and the performance of complex tasks are expected. If these expectations cannot be met due to the appearance on the part of the system's intelligence, this can lead to dissatisfaction on the part of the target audience. The other two appearance styles make it clear to users that this is a technical use, but here there is a double effort, namely that of real social interaction as well as that of technical use. With human appearance and inhuman behavior, there is a risk of appearing threatening, intimidating, and unsettling. However, humanity brings a sense of trust, safety, and comfort. Even though realism seemed scary to some participants, many respondents perceived the realistic avatars as intelligent, professional, knowledgeable, sincere, and authentic.

Finally, it should be emphasized that this is a trend mapping, illustrating that there is no clear consensus of opinion. The evaluation of styles remains a subjective feeling, with each style having its pro and con views. We have provided insights into the framework to be considered for a style assessment, and highlighted some of the advantages as well as disadvantages of each manifestation. In our study context of the terminal and training system, human likeness received the greatest endorsement for confidence and competence. To achieve an increased acceptance of the target audience towards virtual assistants, users can be given the selection option of their preferred avatar type at the beginning of an application, so that they can communicate with their individually most appealing character.

Before choosing an avatar type, ask yourself: Who is your target audience? What are the application requirements (simple, complex, cognitive-analytical, social-emotional, easy language/ sign language, etc.)? What do you want your avatar to represent (fun, approachability, intelligence, etc.)? Do you want a specific type or should your target audience be able to choose?

Do you want to appeal to your target audience with an appropriate virtual appearance? Learn more and contact us:

Your Charamel Team/ Ingmar Dorp
Head of Marketing & Sales
View calendar and schedule a telephone/ online meeting!
Fon: +49 221 33664 25

Charamel GmbH
Aachener Str. 60-62
50674 Cologne, Germany

What is your opinion about the cartoon, human-like and/or photorealistic style? What do you think about the presented findings? We look forward to your feedback!


Our research on end-user acceptance of humanoid avatar styles (cartoon, human-like, realistic) in digital terminal systems (e.g., train station information) and training systems (e.g., eLearning) found that human-like is preferred over realism and cartoon appearance. Cartoon style seems lively and casual, but unsuitable for serious topics. Realism comes across as competent and familiar, but can lead to false expectations of actual capabilities. Human likeness comes across as believable and reliable as an appealing middle ground. Since these are subjective feelings, the selection option of a preferred avatar type can provide a satisfactory solution.