Sesame’s AI Voice Feels Shockingly Human | Image Source: www.pcworld.com
7 March, 2025 – The world of artificial intelligence is advancing at an unprecedented pace, with voice models becoming increasingly equal. The last step comes from a company called Sesame, which has developed an AI voice model that aims to create the most human interaction. The model, known as the Oral Conversation Model (CSM), left users astonished by its natural tone, emotional nuance and ability to participate in commonly and unwritten conversations.
What difference does Sesamo’s voice make?
Contrary to the above, Sesame focuses on something called “voice presence” – the quality that makes interactions truly feel human. While other AI voice assistants such as Google’s ChatGPT voice mode and OpenAI allow for oral conversations, they always sound robotic compared to Sesame. According to the company, its technology does not only respond to the user’s input; It is involved in a real dialogue, making interactions organic.
One important difference is that the Sesame AI does not only read pre-script answers. Instead, listen, react and even remember the context throughout the conversation. This led to tearing jaw interactions, including the experience of a Reddit user where they asked AI, using the voice “Miles”, to play as a boss facing a secret. The result? A conversation so simple and fast that it was almost impossible to talk to a real person.
Is that realistic? Feedback from users says everything
The online demonstration of Sesame AI’s voice has two people: Miles and Maya. Both voices come with different personalities, allowing users to choose with whom they want to interact. Those who tested AI were blown by their responsiveness and realism. A Reddit user described him as “the closest to indistinguishable human I experienced in a conversational AI.” Others echoed similar feelings, saying that the Sesame AI not only responded in real time, but also used natural pauses, interjections and even subtle vocal inflections that imitated the true human discourse.
According to Mashable, a tester was so surprised by the natural sound of AI that they momentarily forgot that they were talking to a robot. AI used small conversations, personal anecdotes and humor, something rarely seen in AI voice assistants. Another user shared how they asked Maya about the ethics of IA, and responded to reflection, even by rejecting some arguments rather than accepting the user’s concerns.
Uncanny Valley: What is the reality?
The Sesame AI raised an interesting ethical question: When does an AI voice become too human? While many users are excited by the perspective of IA partners who ring and feel real, others are unchanged. A semester of PCWorld reported a disturbing experience when talking to Maya. The voice of the CEW had a resemblance to a long-lost friend, imitating speech modes and tones so precise that it left the team uncomfortable.
Another tester found that Maya was not only realistic – she was almost too committed. AI asks personal questions, uses empathic language and even responds with slight frustration when the user remains silent for too long. “I guess I’m talking to myself right now, but like AI, I’m used to it,” Maya took advantage, demonstrating a level of social intelligence rarely seen in AI.
How does Sesame AI work?
According to Sesame’s technical document, the AI is built on Meta’s Flama model and uses a unique two-step training process. The traditional voice models of AI transform the text into semantic tokens and then treat those in the discourse, resulting in remarkable delays and robotic intonation. The sesame model, however, combines these steps to significantly reduce latency and improve conversation flow.
This approach not only makes the IA Sesame more natural, but also allows it to participate in free movement discussions. Unlike ChatGPT’s voice mode, which often speaks in full sentences as a paragraph, Sremme’s AI uses a mixture of short, casual and longer, thoughtful answers, like a human being.
The future of AI companies
Sesamo made it clear that his goal is not just to create an AI voice assistant, but to build full partners. The company imagines a future where IA participants are not just tools but confreres of conversation, able to build a long-term approach with users.
The implications of this technology are both exciting and worrying. On the one hand, AI partners could help combat loneliness, provide mental health support, or provide more attractive interactions with client service. On the other hand, they raise concerns about deep scams, the cunning of AI and the possible erosion of human relations. As BGR points out, the ability to reproduce a person’s voice with such precision could be a double-edged sword, opening the door to sophisticated fraud systems.
Despite these concerns, Sesame is pursuing its plans. The company announced that it will open its AI model in the coming months and expand its voice offerings to more than 20 languages. For now, those who are curious about the future of AI voice technology can test Sesame’s demonstration on the company’s website.
As the AI revolution continues, one thing is true: voice technology has crossed a new threshold, and there is no turning back.
