The world of AI voice assistants has seen a lot of development in recent years, but a new release from AI startup Sesame is making waves for how natural and engaging it feels. After testing Sesame’s Conversational Speech Model (CSM), it became clear that this is a major leap forward. The voice assistant from Sesame left a strong impression, especially when compared to ChatGPT’s voice mode, which felt robotic and disconnected in comparison.
A Conversation That Feels Real
Sesame’s voice assistant, available as a free demo, offers two voices: Maya and Miles. Since its launch, users have been blown away by how natural the assistant sounds. In fact, many have claimed that it’s "about as close to indistinguishable from a human" as they’ve ever experienced in a conversational AI.
During an interaction with Maya, the conversation felt remarkably human. Maya used a natural cadence, interjecting with words like “you know” and “hm” along with non-verbal sounds like tongue clicks and inhales. But what stood out the most was her ability to engage. Maya didn’t wait for me to start the conversation; instead, she immediately asked how my morning was going. It made the whole interaction feel more like a genuine dialogue, not just an exchange of information.
A Starker Difference from ChatGPT
When compared to ChatGPT’s voice mode, Sesame’s assistant feels more fluid and dynamic. ChatGPT’s voice responses often come off as formal and robotic, responding in long paragraphs that can feel like they’ve been lifted from a script. ChatGPT’s voice mode waits for the user to start, and while it offers some conversational flow, it still feels more like a tool rather than a partner in dialogue.
For instance, when discussing concerns about AI companions becoming too human-like, Maya gave a thoughtful response, suggesting that AI should act as a companion, not a replacement for human relationships. In contrast, ChatGPT’s response was far more generic, merely stating the importance of balancing technology with human interactions. It felt less like a conversation and more like a canned answer.
Impressive, but Not Perfect
Despite its impressive abilities, Sesame's voice assistant isn’t without its flaws. Maya’s voice occasionally glitched, and the syntax wasn’t always perfect. For example, she once said, "It's a heavy talk that come," which was awkward but didn't significantly detract from the experience. These minor imperfections aside, the interaction still felt far more natural than anything I've encountered with ChatGPT.
A Leap Forward in AI Conversations
Sesame’s CSM is based on Meta’s Llama model, a similar approach to what OpenAI uses for its voice mode. However, the performance of Sesame’s model is a clear step up in terms of creating a more engaging, fluid conversation. The assistant feels more like a real person than a bot, which is both impressive and, to be honest, a bit unnerving.
While the Sesame demo is just a taste of what’s to come, it’s clear that the company is making significant strides in AI-driven conversations. The CSM has the potential to change how we interact with voice assistants, moving beyond simple task execution to something that feels far more human-like.
Sesame plans to open-source the model in the coming months, with plans to expand it to over 20 languages. This could open up a new era of AI-driven conversations that feel more genuine, pushing the boundaries of what we expect from our digital interactions. For now, the demo has set the bar high—so high, in fact, that it makes ChatGPT’s voice mode feel a bit outdated by comparison.
Need a business app or interested in our reseller program? Reach out!