ChatGPT is revolutionizing the world. It can generate texts, answer questions and even conduct written dialogs, just like a human. It is based on a large language model, GPT, developed by OpenAI and makes work easier for people in many ways. Whether at university, at work or in everyday life, the chatbot uses automated responses to help its users and save them time.

Now OpenAI has added to the many functions and gone one better: With the AI Voice Mode.

It enables users to talk to the AI program and receive a verbal response. This opens up new possibilities for ChatGPT users – and above all a new closeness, as it imitates a human-like conversation.

The following blog article takes a closer look at how Voice Mode works, what advantages it has and what concerns there are.

2.  What is Voice Mode?

ChatGPT’s “Voice Mode” brings a new level of interaction to the AI world: users can have a spoken conversation with the computer. While you can record yourself talking, ChatGPT listens and responds in spoken form. This brings the use of AI one step closer to human interaction. There are two versions that can be used:

Standard Voice Mode: All accounts use the standard voice function with a time limit. This involves transcribing the voice recordings and sending them to the corresponding AI models to generate a response.

Advanced Voice Mode: This voice function is only available for users with a paid subscription and as a limited version for users without a subscription. This includes the ability to generate audio to create natural conversations in real time. Non-verbal cues, such as the speed of speech, are perceived and emotions can be used to react. Users also have the option of choosing from nine different voices that reflect different characters:

1. Light-hearted and versatile
2. Lively and serious
3. Calm and direct
4. Self-confident and optimistic
5. Open and optimistic
6. Cheerful and open
7. Smart and relaxed
8. Calm and affirmative
9. Bright and curious

3.  How do I activate voice mode?

The new voice function is available both in the app and on the website and can be activated at any time.

In the app, there is a language icon at the bottom right of the screen. When users click on this, they are redirected to a screen with a blue or black sphere, depending on which model is being used. You can switch your own microphone on and off and exit the function at any time.

• The voice model works very similarly on the website. However, users must first give the browser access to the microphone so that it can be used. When using it for the first time, users can also select the computer’s voice.

4.  What distinguishes Voice Mode from Siri and co.?

The real-time conversations and the different characters that the ChatGPT language model can take on create a human-like interaction. At first glance, this doesn’t sound all that new. Voice models such as Alexa and Siri have been able to be controlled simply by speaking for years. But at second glance, ChatGPT’s Voice Mode is very different:

Voice Mode, however, is able to conduct coherent, dynamic conversations. These functions are helpful in various situations:

However, the Voice Mode is able to conduct connected, dynamic conversations. These functions are helpful in various situations:

• As an interactive learning aid, for example, to prepare for an interview or an oral exam.
• To practise languages and thus improve pronunciation, fluency and vocabulary.
• To answer spontaneous questions on the go without having to type.
• For people with physical disabilities, for example a visual impairment.

5.  What happens to my personal data?

However, in addition to the many advantages, the new voice function also offers the risk of sharing private information with the computer – more than is already the case. This can also worry users. OpenAI comments on this and explicitly states that voice recordings will remain stored for as long as the chat history is available (Voice Mode FAQ, n.d.). As soon as users delete a chat, the content is also removed from the platform within 30 days. However, this does not apply in cases of security or for legal reasons, where the data is still secured by OpenAI. Users can also release their data so that it can be used to train the AI.

To go into more detail about the voice files, each audio recording is transcribed to generate a response. Only the text is saved, but the voice file is deleted unless users have enabled it to train the AI. This option must be actively enabled by users in the settings and they can select which information is used. Audio or video files must be actively switched on again for this, other data such as transcripts or uploaded images are used automatically. This enables users to secure the personal information they share with the AI, at least up to a certain point.

Conclusion: Voice Mode as AI or conversation partner?

All in all, it can be said that ChatGPT has raised the bar for AI systems enormously with Voice Mode.

The function to hold coherent, dynamic conversations simulates a human conversation, which can be helpful in a wide variety of situations. At the same time, the new function also poses a risk: People are opening themselves up to computers more than ever before. And OpenAI can use this data, for example to further train the AI. Users therefore need to be aware of what they share with ChatGPT and what their data can be used for. However, with the option to protect personal data, OpenAI enables a certain level of privacy and the option to restrict the AI’s access to personal information.

One thing is certain: Voice Mode is a significant step towards natural communication with AI. This function brings humans and computers another step closer.